Channel Catalogue
This catalogue is generated by calling each compiled channel’s descriptor()
and service-route declarations. It includes every channel in the SDK’s channels
feature group. A deployed binary may enable a smaller set.
Capabilities describe implemented optional interfaces, not live account access or a matrix of supported protocol operations. Configuration placeholders are examples, not runtime defaults. Vendor service routes require a provider or namespace mount; access also depends on caller role and service view. A listed route may be restricted or refused, as indicated by its classification. See Providers & Credentials for setup.
GPROXY 4.1.10. Download JSON.
Shared provider configuration
| Key | Kind | Description |
|---|---|---|
forward_thinking | Boolean | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. |
credential_strategy | String | How the host picks among this provider's credentials. |
allowed_headers | HTTP headers | Additional client headers to forward, unioned with the global allow-list and channel defaults. |
Google AI Studio aistudio
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Gemini API origin; defaults to https://generativelanguage.googleapis.com. Provider column, not config JSON. | https://generativelanguage.googleapis.com |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Antigravity (Code Assist) antigravity
Login flows: Authorization code (PKCE). Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Code Assist origin; defaults to https://daily-cloudcode-pa.googleapis.com. Provider column, not config JSON. | https://daily-cloudcode-pa.googleapis.com |
authorize_url | String | No | Google OAuth authorization endpoint for the browser step. | https://accounts.google.com/o/oauth2/v2/auth |
token_url | String | No | Google OAuth token endpoint used by the code exchange and by refresh. | https://oauth2.googleapis.com/token |
headers | HTTP headers | No | Static headers added to every Code Assist request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
AWS Bedrock aws_bedrock
Login flows: API key. Manual credential kinds: api_key, aws_access_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
region | String | No | AWS region for both Bedrock planes and for the SigV4 credential scope; defaults to us-east-1. | — |
base_url | String | No | Origin replacing bedrock-runtime.{region}.amazonaws.com, for a VPC endpoint or a gateway. Provider column, not config JSON. | https://bedrock-runtime.{region}.amazonaws.com |
control_base_url | String | No | Origin replacing bedrock.{region}.amazonaws.com, which serves the foundation-model directory. | https://bedrock.{region}.amazonaws.com |
anthropic_version | String | No | The anthropic_version written into every Anthropic request body; defaults to 2026-10-21. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in an Anthropic model's body into cache_control. The strings are stripped either way. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in a Chat Completions body (non-Anthropic models) into prompt_cache_breakpoint. | — |
fallback_mode | String | No | Claude refusal fallback: off (default), default (recommended model), or models. | — |
fallback_models | JSON | No | Ordered fallback model IDs; duplicates and the primary model are skipped, at most three. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Microsoft Azure azure
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Complete origin of the Azure endpoint, for a Foundry project, a private endpoint or an API Management front door. Provider column, not config JSON. Replaces `resource`. | https://{resource}.openai.azure.com |
resource | String | No | Azure OpenAI resource name; the origin becomes https://{resource}.openai.azure.com. Required unless the provider has a base_url. | — |
api_version | String | No | api-version query value. Optional on the v1 surface; required when `deployment` is set. | — |
deployment | String | No | Deployment name. Set it to address the deployment-scoped layout /openai/deployments/{deployment}/... instead of the version-less v1 surface. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Claude-dialect body into cache_control. The strings are stripped either way. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in an OpenAI Chat or Responses body into prompt_cache_breakpoint. | — |
fallback_mode | String | No | Claude refusal fallback: off (default), default (recommended model), or models. | — |
fallback_models | JSON | No | Ordered fallback model IDs; duplicates and the primary model are skipped, at most three. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Claude API (Anthropic) claudeapi
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Anthropic API origin; defaults to https://api.anthropic.com. Provider column, not config JSON. | https://api.anthropic.com |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Messages or count_tokens body into cache_control. The strings are stripped either way. | — |
fallback_mode | String | No | Server-side fallback for Messages requests that name none: off (default), default, or models. | — |
fallback_models | JSON | No | The chain fallback_mode `models` installs; at most three are sent. | — |
quota_base_url | String | No | Origin of the organization cost report when it differs from base_url. | https://api.anthropic.com |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Claude Code (Claude.ai) claudecode
Login flows: Authorization code (PKCE). Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | Yes |
| Credential refresh | Yes |
| Vendor service routes | Yes |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Anthropic API origin; defaults to https://api.anthropic.com. Provider column, not config JSON. | https://api.anthropic.com |
authorize_url | String | No | OAuth authorization endpoint for the browser step. | https://claude.com/cai/oauth/authorize |
token_url | String | No | OAuth token endpoint used by the code exchange and by refresh. | https://platform.claude.com/v1/oauth/token |
claude_ai_url | String | No | The claude.ai origin the cookie initialization talks to. | https://claude.ai |
headers | HTTP headers | No | Static headers added to every backend request. | — |
fallback_mode | String | No | Server-side fallback for Messages requests that name none: off (default), default, or models. | — |
fallback_models | JSON | No | The ordered chain fallback_mode models installs; at most three are sent. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Messages or count_tokens body into cache_control. | — |
low_priority | Boolean | No | Accept lower-priority offers after a full 5h window and wait for capacity like Claude Code. Off by default; the server decides eligibility and budget. | — |
extended_usage | Boolean | No | Request extended usage grace for tool-result continuations unless paid extra usage is enabled. Off by default; the server decides whether grace is available. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Vendor service routes (76)
| Method | Path | Transport | Access classification | Idempotent |
|---|---|---|---|---|
GET | /v1/mcp_servers | HTTP | Restricted | Yes |
GET | /mcp-registry/v0/servers | HTTP | Catalogue | Yes |
GET | /api/directory/servers | HTTP | Catalogue | Yes |
GET | /api/hello | HTTP | Catalogue | Yes |
GET | /api/hello/{name} | HTTP | Catalogue | Yes |
GET | /api/oauth/profile | HTTP | Identity | Yes |
POST | /api/oauth/validate | HTTP | Identity | Yes |
GET | /api/oauth/claude_cli/roles | HTTP | Identity | Yes |
POST | /api/oauth/claude_cli/create_api_key | HTTP | Refused | No |
GET | /api/oauth/cri | HTTP | Identity | Yes |
GET | /api/claude_cli_profile | HTTP | Identity | Yes |
POST | /api/auth/trusted_devices | HTTP | Restricted | No |
GET | /api/claude_cli/bootstrap | HTTP | Identity | Yes |
GET | /api/oauth/usage | HTTP | Usage | Yes |
GET | /api/claude_code/policy_limits | HTTP | Settings | Yes |
GET | /api/claude_code/settings | HTTP | Settings | Yes |
GET | /api/claude_code/skills | HTTP | Settings | Yes |
GET | /api/claude_code/notification/preferences | HTTP | Settings | Yes |
GET | /api/claude_code/organizations/metrics_enabled | HTTP | Settings | Yes |
POST | /api/claude_code/metrics | HTTP | Telemetry | No |
GET | /api/organization/claude_code_first_token_date | HTTP | Settings | Yes |
GET | /api/claude_code_penguin_mode | HTTP | Settings | Yes |
GET | /api/claude_code_grove | HTTP | Settings | Yes |
GET | /api/oauth/account/settings | HTTP | Settings | Yes |
PATCH | /api/oauth/account/settings | HTTP | Settings | Yes |
POST | /api/oauth/account/grove_notice_viewed | HTTP | Settings | Yes |
POST | /api/claude_cli_feedback | HTTP | Telemetry | No |
POST | /api/claude_code_shared_session_transcripts | HTTP | Telemetry | No |
POST | /api/oauth/file_upload | HTTP | Resource: claude:file / Create | No |
GET | /api/oauth/files/{file_uuid}/content | HTTP | Resource: claude:file / Read / act on an item | Yes |
POST | /api/oauth/organizations/{org}/mcp/connectors/search | HTTP | Settings | Yes |
POST | /api/oauth/organizations/{org}/mcp/connectors/suggest | HTTP | Settings | Yes |
POST | /api/oauth/organizations/{org}/mcp/connectors/list | HTTP | Settings | Yes |
POST | /api/oauth/organizations/{org}/plugins/search | HTTP | Resource: claude:plugin / List | Yes |
GET | /api/oauth/organizations/{org}/plugins/list-plugins | HTTP | Resource: claude:plugin / List | Yes |
GET | /api/oauth/organizations/{org}/plugins/{plugin_id}/download | HTTP | Resource: claude:plugin / Read / act on an item | Yes |
POST | /api/oauth/organizations/{org}/skills/search | HTTP | Resource: claude:skill / List | Yes |
GET | /api/oauth/organizations/{org}/skills/list-skills | HTTP | Resource: claude:skill / List | Yes |
GET | /api/oauth/organizations/{org}/skills/{skill_id}/download | HTTP | Resource: claude:skill / Read / act on an item | Yes |
POST | /api/oauth/organizations/{org}/plugin_ratings | HTTP | Telemetry | No |
POST | /api/oauth/organizations/{org}/plugin_ratings/appearances | HTTP | Telemetry | No |
POST | /api/oauth/organizations/{org}/plugin_ratings/appearances/outcome | HTTP | Telemetry | No |
DELETE | /api/oauth/organizations/{org}/plugin_ratings | HTTP | Telemetry | Yes |
GET | /api/oauth/organizations/{org}/sync/github/auth | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/payment_method | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/prepaid/bundles | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/prepaid/credits | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/prepaid/commits/{commit_id} | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/admin_requests/eligibility | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/admin_requests/me | HTTP | Restricted | Yes |
POST | /api/oauth/organizations/{org}/admin_requests | HTTP | Restricted | No |
POST | /api/oauth/organizations/{org}/billing/tax_rate | HTTP | Restricted | Yes |
POST | /api/oauth/organizations/{org}/claude_code/pro_trial | HTTP | Restricted | No |
POST | /api/oauth/organizations/{org}/contracts/prepaid/credits | HTTP | Restricted | No |
PUT | /api/oauth/organizations/{org}/contracts/auto_reload_settings | HTTP | Restricted | Yes |
GET | /api/oauth/organizations/{org}/overage_credit_grant | HTTP | Restricted | Yes |
POST | /api/oauth/organizations/{org}/overage_credit_grant | HTTP | Restricted | No |
PUT | /api/oauth/organizations/{org}/overage_spend_limit | HTTP | Restricted | Yes |
POST | /api/oauth/organizations/{org}/setup_overage_billing | HTTP | Restricted | No |
GET | /api/oauth/organizations/{org}/referral/eligibility | HTTP | Restricted | Yes |
POST | /api/oauth/organizations/{org}/referral/redemptions | HTTP | Restricted | No |
POST | /api/organizations/{org}/reset_rate_limits | HTTP | Restricted | No |
GET | /api/organizations/{org}/model_selector/{selector} | HTTP | Restricted | Yes |
GET | /api/organizations/{org}/claude_code/onboarding | HTTP | Restricted | Yes |
POST | /api/organizations/{org}/claude_code/onboarding | HTTP | Restricted | No |
GET | /api/organizations/{org}/cowork/remote_devices | HTTP | Restricted | Yes |
GET | /api/organizations/{org}/marketplace/plugin | HTTP | Restricted | Yes |
GET | /api/organizations/{org}/projects/{*path} | HTTP | Restricted | Yes |
GET | /api/desktop/{*path} | HTTP | Catalogue | Yes |
GET | /api/frame/{*path} | HTTP | Restricted | Yes |
POST | /api/frame/{*path} | HTTP | Restricted | No |
GET | /api/{*path} | HTTP | Prefix fallback | Yes |
POST | /api/{*path} | HTTP | Prefix fallback | No |
PUT | /api/{*path} | HTTP | Prefix fallback | No |
PATCH | /api/{*path} | HTTP | Prefix fallback | No |
DELETE | /api/{*path} | HTTP | Prefix fallback | No |
Claude Web (claude.ai session) claudeweb
Login flows: None. Manual credential kinds: cookie.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | claude.ai origin; defaults to https://claude.ai. Provider column, not config JSON. | https://claude.ai |
prompt | String | No | Text placed before the flattened conversation in every prompt. | — |
timezone | String | No | The timezone the web request declares; defaults to UTC. | — |
endpoints | JSON | No | Per-step URL overrides keyed claudeweb_upload, claudeweb_completion, claudeweb_bootstrap and friends; the organization and conversation placeholders are substituted. | — |
headers | HTTP headers | No | Static headers added to every claude.ai request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Cline cline
Login flows: API key / Device code. Manual credential kinds: api_key, oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Cline API origin, `/api/v1` included; defaults to https://api.cline.bot/api/v1. Provider column, not config JSON. | https://api.cline.bot/api/v1 |
device_authorization_url | String | No | Where the device login starts; defaults to WorkOS's device authorization endpoint. | https://api.workos.com/user_management/authorize/device |
token_url | String | No | Where the device login polls for the WorkOS token pair. | https://api.workos.com/user_management/authenticate |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Cloudflare AI Gateway cloudflare_ai_gateway
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Cloudflare REST API origin; the account path is appended. Provider column, not config JSON. | https://api.cloudflare.com |
headers | HTTP headers | No | Static headers added to every upstream request, such as cf-aig-* switches. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
OpenAI Codex (ChatGPT) codex
Login flows: Authorization code (PKCE) / Device code. Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | Yes |
| Credential refresh | Yes |
| Vendor service routes | Yes |
| WebSocket messages | Yes |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Codex backend origin and prefix; defaults to https://chatgpt.com/backend-api/codex. Provider column, not config JSON. | https://chatgpt.com/backend-api/codex |
issuer | String | No | OAuth issuer used for login and refresh; defaults to https://auth.openai.com. | https://auth.openai.com |
originator | String | No | The `originator` the backend sees; defaults to codex_cli_rs. | codex_cli_rs |
headers | HTTP headers | No | Static headers added to every backend request. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in a Responses body into prompt_cache_breakpoint. | — |
synthesize_cli_identity | Boolean | No | Give non-CLI clients the CLI's session, thread, window and turn identity headers. On by default. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Vendor service routes (76)
| Method | Path | Transport | Access classification | Idempotent |
|---|---|---|---|---|
GET | /v1/user-auth-credential/whoami | HTTP | Identity | Yes |
GET | /backend-api/wham/accounts/check | HTTP | Identity | Yes |
POST | /backend-api/wham/accounts/send_add_credits_nudge_email | HTTP | Restricted | No |
GET | /backend-api/wham/profiles/me | HTTP | Identity | Yes |
GET | /backend-api/wham/settings/user | HTTP | Settings | Yes |
GET | /backend-api/wham/workspace-messages | HTTP | Settings | Yes |
GET | /backend-api/accounts/verified_access | HTTP | Identity | Yes |
GET | /backend-api/accounts/{account_id}/settings | HTTP | Identity | Yes |
GET | /backend-api/wham/usage | HTTP | Usage | Yes |
POST | /backend-api/wham/usage/thread_usage/query_v2 | HTTP | Usage | Yes |
GET | /backend-api/wham/usage/plan_limit_history | HTTP | Usage | Yes |
GET | /backend-api/wham/usage/daily-token-usage-breakdown | HTTP | Usage | Yes |
GET | /backend-api/wham/usage/credit-usage-events | HTTP | Usage | Yes |
GET | /backend-api/wham/usage/daily-workspace-user-token-usage-breakdown | HTTP | Usage | Yes |
GET | /backend-api/wham/usage/daily-workspace-user-credit-usage | HTTP | Usage | Yes |
GET | /backend-api/wham/analytics/daily-workspace-usage-counts | HTTP | Usage | Yes |
GET | /backend-api/wham/analytics/daily-plugin-usage-metrics | HTTP | Usage | Yes |
GET | /backend-api/wham/analytics/daily-skill-usage-metrics | HTTP | Usage | Yes |
POST | /backend-api/wham/usage/thread_usage/query | HTTP | Usage | Yes |
POST | /backend-api/wham/usage/thread-estimates/query | HTTP | Usage | Yes |
GET | /backend-api/wham/rate-limit-reset-credits | HTTP | Usage | Yes |
POST | /backend-api/wham/rate-limit-reset-credits/consume | HTTP | Usage | No |
POST | /backend-api/analytics/codex/turn-costs | HTTP | Telemetry | No |
GET | /backend-api/wham/tasks | HTTP | Resource: codex:task / List | Yes |
POST | /backend-api/wham/tasks | HTTP | Resource: codex:task / Create | No |
GET | /backend-api/wham/tasks/list | HTTP | Resource: codex:task / List | Yes |
GET | /backend-api/wham/tasks/{task_id} | HTTP | Resource: codex:task / Read / act on an item | Yes |
GET | /backend-api/wham/tasks/{task_id}/turns/{turn_id}/sibling_turns | HTTP | Resource: codex:task / Read / act on an item | Yes |
GET | /backend-api/wham/environments | HTTP | Resource: codex:environment / List | Yes |
GET | /backend-api/wham/environments/by-repo/{host}/{owner}/{repo} | HTTP | Resource: codex:environment / List | Yes |
GET | /backend-api/wham/config/bundle | HTTP | Settings | Yes |
GET | /backend-api/wham/app/appcast | HTTP | Catalogue | Yes |
GET | /backend-api/wham/agent-identities/jwks | HTTP | Catalogue | Yes |
POST | /backend-api/v1/agent/register | HTTP | Restricted | No |
POST | /backend-api/v1/agent/{agent_runtime_id}/task/register | HTTP | Restricted | No |
GET | /backend-api/wham/remote/control/server | WebSocket | Restricted | Yes |
POST | /backend-api/wham/remote/control/server/enroll | HTTP | Resource: codex:remote_server / Create | No |
POST | /backend-api/wham/remote/control/server/refresh | HTTP | Restricted | No |
POST | /backend-api/wham/remote/control/server/pair | HTTP | Restricted | No |
GET | /backend-api/wham/remote/control/server/pair/status | HTTP | Restricted | Yes |
GET | /backend-api/wham/remote/control/environments/{environment_id}/clients | HTTP | Resource: codex:remote_server / List | Yes |
DELETE | /backend-api/wham/remote/control/environments/{environment_id}/clients/{client_id} | HTTP | Resource: codex:remote_server / Delete | No |
POST | /backend-api/codex/alpha/search | HTTP | Catalogue | Yes |
GET | /backend-api/codex/models | HTTP | Catalogue | Yes |
POST | /backend-api/codex/alpha/{namespace}/v2/{tool} | HTTP | Restricted | No |
POST | /backend-api/codex/analytics-events/events | HTTP | Telemetry | No |
POST | /backend-api/codex/realtime/calls | HTTP | Catalogue | No |
POST | /backend-api/files | HTTP | Resource: codex:file / Create | No |
POST | /backend-api/files/{file_id}/uploaded | HTTP | Resource: codex:file / Read / act on an item | No |
GET | /backend-api/ps/plugins/list | HTTP | Catalogue | Yes |
GET | /backend-api/ps/plugins/search | HTTP | Catalogue | Yes |
GET | /backend-api/ps/plugins/installed | HTTP | Resource: codex:plugin / List | Yes |
GET | /backend-api/ps/plugins/suggested/codex | HTTP | Catalogue | Yes |
GET | /backend-api/ps/plugins/workspace/created | HTTP | Resource: codex:plugin / List | Yes |
GET | /backend-api/ps/plugins/workspace/shared | HTTP | Resource: codex:plugin / List | Yes |
GET | /backend-api/ps/plugins/{plugin_id} | HTTP | Catalogue | Yes |
POST | /backend-api/ps/plugins/{plugin_id}/install | HTTP | Resource: codex:plugin / Create | No |
POST | /backend-api/ps/plugins/{plugin_id}/uninstall | HTTP | Resource: codex:plugin / Delete | No |
PUT | /backend-api/ps/plugins/{plugin_id}/shares | HTTP | Resource: codex:plugin / Read / act on an item | Yes |
POST | /backend-api/ps/apps/batch | HTTP | Restricted | Yes |
GET | /backend-api/ps/mcp | HTTP | Restricted | Yes |
POST | /backend-api/ps/mcp | HTTP | Restricted | No |
DELETE | /backend-api/ps/mcp | HTTP | Restricted | Yes |
GET | /backend-api/plugins/featured | HTTP | Catalogue | Yes |
GET | /backend-api/plugins/export/curated | HTTP | Catalogue | Yes |
GET | /backend-api/connectors/directory/list | HTTP | Catalogue | Yes |
GET | /backend-api/connectors/directory/list_workspace | HTTP | Restricted | Yes |
POST | /backend-api/public/plugins/workspace | HTTP | Restricted | No |
POST | /backend-api/public/plugins/workspace/upload-url | HTTP | Restricted | No |
POST | /backend-api/public/plugins/workspace/{plugin_id} | HTTP | Restricted | No |
DELETE | /backend-api/public/plugins/workspace/{plugin_id} | HTTP | Restricted | No |
GET | /backend-api/{*path} | HTTP | Prefix fallback | Yes |
POST | /backend-api/{*path} | HTTP | Prefix fallback | No |
PUT | /backend-api/{*path} | HTTP | Prefix fallback | No |
PATCH | /backend-api/{*path} | HTTP | Prefix fallback | No |
DELETE | /backend-api/{*path} | HTTP | Prefix fallback | No |
Command Code commandcode
Login flows: API key / Authorization code (PKCE). Manual credential kinds: api_key, oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Command Code API origin, without /alpha or /provider. | https://api.commandcode.ai |
headers | HTTP headers | No | Static upstream headers. | — |
project_slug | String | No | Project identity sent to the CLI API; defaults to gproxy. | — |
server_config | JSON | No | CLI workspace context sent as config. Defaults to an empty object. | — |
zero_data_retention | Boolean | No | Send x-cmd-zdr: 1, matching CMD_ZDR=1 in the CLI. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
GitHub Copilot CLI copilotcli
Login flows: Device code. Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Copilot inference origin; defaults to the one the seat implies. Provider column, not config JSON. | — |
account_type | String | No | `individual`, `business` or `enterprise`. Decides the inference origin when base_url names none. | — |
github_api_url | String | No | Where the Copilot token is minted and the seat probed; defaults to https://api.github.com. | https://api.github.com |
device_authorization_url | String | No | Where the device login starts; defaults to GitHub's device code endpoint. | https://github.com/login/device/code |
token_url | String | No | Where the device login polls for the GitHub token. | https://github.com/login/oauth/access_token |
scope | String | No | Scope the device login asks for; defaults to read:user. | — |
vscode_version | String | No | Editor release named in editor-version on GitHub calls. | — |
headers | HTTP headers | No | Static headers added to every inference request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Custom (API key) custom
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | Yes | Upstream origin, optionally with a path prefix; the native request path is appended. Provider column, not config JSON. | — |
dialects | JSON | No | Wire dialects the upstream accepts natively. Empty accepts openai, openai_chat, claude and gemini. | — |
auth_header | String | No | Header carrying the API key instead of each family's own, e.g. `api-key`. | — |
auth_prefix | String | No | Prefix written before the key when auth_header is set. | — |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Claude-dialect body into cache_control. The strings are stripped either way. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in an OpenAI Chat or Responses body into prompt_cache_breakpoint. | — |
fallback_mode | String | No | Claude refusal fallback: off (default), default (recommended model), or models. | — |
fallback_models | JSON | No | Ordered fallback model IDs; duplicates and the primary model are skipped, at most three. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Alibaba DashScope dashscope
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin; defaults to https://dashscope.aliyuncs.com. The compatibility prefixes are appended by the channel. Provider column, not config JSON. | https://dashscope.aliyuncs.com |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
DeepSeek deepseek
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin; defaults to https://api.deepseek.com. Provider column, not config JSON. | https://api.deepseek.com |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Devin (Windsurf, server.codeium.com) devin
Login flows: Authorization code (PKCE). Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Connect-RPC origin; defaults to https://server.codeium.com. Provider column, not config JSON. | https://server.codeium.com |
authorize_url | String | No | Devin browser PKCE authorization endpoint. | https://app.devin.ai/auth/cli/continue |
token_url | String | No | Devin authorization-code exchange endpoint. | https://api.devin.ai/auth/cli/token |
client_name | String | No | ClientMetadata #1 and #12; defaults to the CLI's own name. | — |
client_version | String | No | ClientMetadata #2 and #7; defaults to the captured CLI build. | — |
locale | String | No | ClientMetadata #4; defaults to en. | — |
os | String | No | ClientMetadata #5; defaults to windows, as the capture carried. | — |
models | JSON | No | Operator-defined model aliases and additional selectors. Discovery uses the live upstream catalogue; unmapped selectors pass through unchanged and are never downgraded to another model. | — |
max_tokens | Integer | No | CompletionConfig #2 when the request names no cap. | — |
context_window | Integer | No | CompletionConfig #3. The reference declares this field max_newlines and writes the context window into it; the key is named for the use, not the declaration. Also accepted as max_newlines. | — |
headers | HTTP headers | No | Static headers added to every call. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Gemini CLI (Code Assist) geminicli
Login flows: Authorization code (PKCE). Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Code Assist origin; defaults to https://cloudcode-pa.googleapis.com. Provider column, not config JSON. | https://cloudcode-pa.googleapis.com |
authorize_url | String | No | Google OAuth authorization endpoint for the browser step. | https://accounts.google.com/o/oauth2/v2/auth |
token_url | String | No | Google OAuth token endpoint used by the code exchange and by refresh. | https://oauth2.googleapis.com/token |
headers | HTTP headers | No | Static headers added to every Code Assist request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
GLM API / Z.ai glm
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin: https://open.bigmodel.cn (China) or https://api.z.ai (global). | https://open.bigmodel.cn |
headers | HTTP headers | No | Static upstream headers. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
GLM Coding Plan / Z.ai glmcode
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin: https://open.bigmodel.cn (China) or https://api.z.ai (global). | https://open.bigmodel.cn |
headers | HTTP headers | No | Static upstream headers. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Grok Build (xAI account) grokbuild
Login flows: Device code. Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Chat proxy origin; defaults to https://cli-chat-proxy.grok.com/v1. Provider column, not config JSON. | https://cli-chat-proxy.grok.com/v1 |
media_base_url | String | No | Origin for xAI's own image, speech, transcription and video paths; defaults to https://api.x.ai/v1. | https://api.x.ai/v1 |
usage_base_url | String | No | Origin the billing probe reads; defaults to the chat proxy, which is the only surface that answers it. | {base_url} |
oauth_device_code_url | String | No | Device authorization endpoint; defaults to https://auth.x.ai/oauth2/device/code. | https://auth.x.ai/oauth2/device/code |
oauth_token_url | String | No | Token endpoint used by the device poll and by refresh; defaults to https://auth.x.ai/oauth2/token. | https://auth.x.ai/oauth2/token |
headers | HTTP headers | No | Static headers added to every request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Moonshot (Kimi) kimi
Login flows: API key / Device code. Manual credential kinds: api_key, oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin; defaults to https://api.moonshot.cn for a key and https://api.kimi.com/coding/v1 for a subscription. Provider column, not config JSON. | — |
product | String | No | `platform` or `code`. Says whether this provider's upstream answers Claude Messages natively; read off base_url when unstated. | — |
oauth_host | String | No | Where the device login and refresh talk; defaults to https://auth.kimi.com. | https://auth.kimi.com |
cli_version | String | No | Version announced in user-agent and x-msh-version. | — |
device_name | String | No | Machine name announced in x-msh-device-name; `unknown` when unset. | — |
device_model | String | No | Hardware announced in x-msh-device-model; the build's os and arch when unset. | — |
os_version | String | No | Release announced in x-msh-os-version; the build's target OS when unset. | — |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Kiro (AWS CodeWhisperer) kiro
Login flows: Device code / Authorization code (PKCE). Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Runtime origin, replacing https://runtime.{region}.kiro.dev; also serves the management plane unless management_base_url names one. Provider column, not config JSON. | https://runtime.{region}.kiro.dev |
region | String | No | Region for both Kiro planes and for the AWS OIDC host an Identity Center credential refreshes against; defaults to us-east-1. | us-east-1 |
management_base_url | String | No | Origin replacing https://management.{region}.kiro.dev, which serves the model catalogue and the usage limits. | https://management.{region}.kiro.dev |
auth_base_url | String | No | The Kiro desktop auth host the device login and its refresh talk to. | https://prod.us-east-1.auth.desktop.kiro.dev |
profile_arn | String | No | CodeWhisperer profile ARN for credentials whose login did not report one; a credential's own wins. | — |
login_provider | String | No | Identity provider the device login opens: github or google. | — |
sso_start_url | String | No | The Identity Center portal the authorization code is issued by; defaults to the Builder ID portal. | https://view.awsapps.com/start |
headers | HTTP headers | No | Static headers added to every request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
MiniMax minimax
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | MiniMax API origin. Text accepts a Subscription Key or API Key; H3 video requires a pay-as-you-go API Key. | https://api.minimax.io |
headers | HTTP headers | No | Static upstream headers. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
NVIDIA NIM nvidia
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin, without `/v1`. Provider column, not config JSON. | https://integrate.api.nvidia.com |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
OpenAI openai
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | Yes |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | OpenAI API origin; defaults to https://api.openai.com. Provider column, not config JSON. | https://api.openai.com |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in a Chat Completions or Responses body into prompt_cache_breakpoint. The strings are stripped either way. | — |
quota_base_url | String | No | Origin of the organization cost report when it differs from base_url. | https://api.openai.com |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
OpenCode Go opencodego
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin, `/v1` included. Provider column, not config JSON. | https://opencode.ai/zen/go/v1 |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Claude-dialect body into cache_control. The strings are stripped either way. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in an OpenAI Chat or Responses body into prompt_cache_breakpoint. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
OpenCode Zen opencodezen
Login flows: API key / Device code. Manual credential kinds: api_key, oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin, `/v1` included. Provider column, not config JSON. | https://opencode.ai/zen/v1 |
console_base_url | String | No | Where the device login and the refresh talk. | https://opencode.ai/console |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Claude-dialect body into cache_control. The strings are stripped either way. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in an OpenAI Chat or Responses body into prompt_cache_breakpoint. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
OpenRouter openrouter
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin; defaults to https://openrouter.ai/api. Provider column, not config JSON. | https://openrouter.ai/api |
referer | String | No | HTTP-Referer sent when the client sends none; OpenRouter attributes traffic to it. | — |
title | String | No | X-Title sent when the client sends none. | — |
provider | JSON | No | Default provider routing preferences (order, only, ignore, allow_fallbacks, sort, require_parameters, data_collection, quantizations, max_price), forwarded as the body's `provider`. | — |
model_providers | JSON | No | Routing preferences per upstream model name, each replacing `provider` whole for that model. | — |
usage_accounting | Boolean | No | Ask for usage.include so the reply reports the price charged; on by default. | — |
normalize_service_tier | Boolean | No | Rewrite OpenAI's service_tier `fast` into OpenRouter's `priority`; on by default. | — |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
enable_claude_magic_cache | Boolean | No | Turn a client's magic cache string in a Claude-dialect body into cache_control. The strings are stripped either way. | — |
enable_openai_magic_cache | Boolean | No | Turn a client's magic cache string in an OpenAI Chat or Responses body into prompt_cache_breakpoint. | — |
fallback_mode | String | No | Claude refusal fallback: off (default), default (recommended model), or models. | — |
fallback_models | JSON | No | Ordered fallback model IDs; duplicates and the primary model are skipped, at most three. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
OrcaTerm orcaterm
Login flows: Device code. Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | LightAI origin, without /assistant. Defaults to the selected OrcaTerm region. | https://lightai.cloud.tencent.com |
international | Boolean | No | Use OrcaTerm's international account and LightAI endpoints. Defaults to false. | — |
site_url | String | No | Override the OrcaTerm account origin used for OAuth, cookie initialization and token refresh. | https://orcaterm.com |
oauth_provider | String | No | OAuth login provider: wechat, txcloud, google or github. Defaults to wechat in China and google internationally. | — |
headers | HTTP headers | No | Static upstream headers. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Trae trae
Login flows: Authorization code (PKCE). Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
international | Boolean | No | Use international Trae; defaults to mainland China. Account region selects the international service origin. | — |
base_url | String | No | Chat and model API origin override. Defaults to the selected edition and account region. | — |
auth_base_url | String | No | Authentication and account API origin override. Normally selected from the edition and account region. | — |
ide_version | String | No | Desktop version override. Defaults to the audited edition's version. | — |
headers | HTTP headers | No | Static upstream headers; account authentication and device identity remain credential-owned. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Vercel AI Gateway vercel
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Gateway origin; defaults to https://ai-gateway.vercel.sh. | https://ai-gateway.vercel.sh |
headers | HTTP headers | No | Static upstream headers. | — |
enable_claude_magic_cache | Boolean | No | Place Claude cache breakpoints at magic cache strings. | — |
enable_openai_magic_cache | Boolean | No | Place OpenAI cache breakpoints at magic cache strings. | — |
fallback_mode | String | No | Claude refusal fallback: off (default), default (recommended model), or models. | — |
fallback_models | JSON | No | Ordered fallback model IDs; duplicates and the primary model are skipped, at most three. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Google Vertex AI vertex
Login flows: None. Manual credential kinds: service_account.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
project | String | No | Google Cloud project id. Required unless the service-account key carries its own project_id. | — |
location | String | No | Vertex region, for example us-central1, europe-west4 or global. Decides both the origin host and the method path; credential location overrides this default; defaults to us-central1. | us-central1 |
base_url | String | No | Complete origin, replacing the regional https://{location}-aiplatform.googleapis.com. Provider column, not config JSON. | https://{location}-aiplatform.googleapis.com |
fallback_mode | String | No | Claude refusal fallback: off (default), default (recommended model), or models. | — |
fallback_models | JSON | No | Ordered fallback model IDs; duplicates and the primary model are skipped, at most three. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Google Vertex AI Express vertexexpress
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | No |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
location | String | No | Vertex region, for example us-central1 or global. Credential location overrides this default; defaults to global. Only the origin changes, not the publisher-scoped path. | global |
base_url | String | No | Complete origin, replacing https://aiplatform.googleapis.com. Provider column, not config JSON. Overrides the region-derived origin; Express mode requires no project. | https://aiplatform.googleapis.com |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Volcengine Agent Plan volcengine_agent
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | Yes |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | Yes |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Ark inference origin, without /api/plan or /api/coding. Agent Plan requires its dedicated API key. | https://ark.cn-beijing.volces.com |
edition | String | No | Plan edition: personal (default) or enterprise. Enterprise quota requires credential metadata.seat_id. | — |
control_base_url | String | No | Ark control-plane origin. Model discovery, plan information and quota require access_key_id and secret_access_key in the credential secret. | https://ark.cn-beijing.volcengineapi.com |
speech_base_url | String | No | Agent Plan speech origin for HTTP TTS. | https://openspeech.bytedance.com |
headers | HTTP headers | No | Static upstream headers. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Vendor service routes (1)
| Method | Path | Transport | Access classification | Idempotent |
|---|---|---|---|---|
GET | /api/volcengine/plan | HTTP | Restricted | Yes |
Volcengine Coding Plan volcengine_coding
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | Yes |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Ark inference origin, without /api/plan or /api/coding. Agent Plan requires its dedicated API key. | https://ark.cn-beijing.volces.com |
edition | String | No | Plan edition: personal (default) or enterprise. Enterprise quota requires credential metadata.seat_id. | — |
control_base_url | String | No | Ark control-plane origin. Model discovery, plan information and quota require access_key_id and secret_access_key in the credential secret. | https://ark.cn-beijing.volcengineapi.com |
speech_base_url | String | No | Agent Plan speech origin for HTTP TTS. | https://openspeech.bytedance.com |
headers | HTTP headers | No | Static upstream headers. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
Vendor service routes (1)
| Method | Path | Transport | Access classification | Idempotent |
|---|---|---|---|---|
GET | /api/volcengine/plan | HTTP | Restricted | Yes |
WorkBuddy (Tencent Copilot) workbuddy
Login flows: Device code. Manual credential kinds: oauth.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | Yes |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
international | Boolean | No | Use WorkBuddy AI's international service and login platform. Defaults to false (mainland). | — |
base_url | String | No | Upstream origin override; defaults to https://copilot.tencent.com, or https://www.workbuddy.ai when international is true. Provider column, not config JSON. | — |
ide_version | String | No | Plugin version reported in x-ide-version and in the user agent. | — |
ide_name | String | No | What x-ide-name claims the plugin runs inside; the CLI says CLI. | — |
ide_type | String | No | What x-ide-type claims the plugin runs inside; the CLI says CLI. | — |
agent_intent | String | No | The intent x-agent-intent asks the agent surface for; the CLI says craft. | — |
headers | HTTP headers | No | Static headers added to every request. | — |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |
xAI (Grok) xai
Login flows: API key. Manual credential kinds: api_key.
| Capability | Declared support |
|---|---|
| Paid usage | No |
| Quota query | Yes |
| Quota reset | No |
| Credential refresh | No |
| Vendor service routes | No |
| WebSocket messages | No |
| Key | Kind | Required | Description | Example / placeholder |
|---|---|---|---|---|
base_url | String | No | Upstream origin; defaults to https://api.x.ai. Provider column, not config JSON. | https://api.x.ai |
headers | HTTP headers | No | Static headers added to every upstream request. | — |
quota_api_key | String | No | An xAI management key, which is not the inference key. Without it there is no billing surface to read. | — |
quota_team_id | String | No | The team whose prepaid balance and postpaid limit the probe reads. | — |
quota_base_url | String | No | Management origin; defaults to https://management-api.x.ai. | https://management-api.x.ai |
forward_thinking | Boolean | No | Forward thinking blocks from request history. Defaults to true; disabling removes reasoning content and signatures without disabling model thinking. | — |
credential_strategy | String | No | How the host picks among this provider's credentials. | — |
allowed_headers | HTTP headers | No | Additional client headers to forward, unioned with the global allow-list and channel defaults. | — |