Mission Squad Docs

Appearance

Models ​

Model operations include discovery from configured providers, adding/removing models in your namespace, and simple listing endpoints.

Endpoints ​

GET /v1/models ​

Returns a combined list of your models plus chat-eligible agents.

  • Owned agents appear here only when they are published.
  • Shared agents appear here when they are published and explicitly shared to the authenticated caller.
  • Shared agents use a canonical id of shared/<ownerUsername>/<slug>.

Each entry:

json
{ "id": "my-gpt4", "object": "model", "owned_by": "user" }

Shared agent example:

json
{ "id": "shared/john_doe/customer-support-bot", "object": "agent", "owned_by": "john_doe" }

Example:

ts
await fetch("https://agents.missionsquad.ai/v1/models", {
  headers: { "x-api-key": process.env.MSQ_API_KEY! }
});

GET /v1/modelmap ​

Returns a detailed map keyed by name. Secrets are masked.

ts
await fetch("https://agents.missionsquad.ai/v1/modelmap", {
  headers: { "x-api-key": process.env.MSQ_API_KEY! }
});

POST /v1/core/models — Discover provider models ​

Body:

ts
{
  providerKey: string,
  url?: string,     // optional override
  apiKey?: string   // optional ad-hoc; stored creds take precedence if already configured
}

Notes:

  • If the provider is already configured for your account, stored credentials are used and any apiKey/url in the request body are ignored.

POST /v1/core/add/model — Add a model in your namespace ​

Body:

json
{
  "name": "my-claude",
  "description": "Claude Sonnet 4.6",
  "providerKey": "anthropic",
  "model": "claude-sonnet-4-6",
  "testResponse": true,
  "getAllApiModels": false,
  "extractEmbeddingModels": false,
  "programmaticToolCalling": true
}

Full body fields:

ts
{
  name: string;
  description: string;
  providerKey: string;
  model: string;                    // the provider's model id
  testResponse?: boolean;
  getAllApiModels?: boolean;
  extractEmbeddingModels?: boolean;
  programmaticToolCalling?: boolean; // accepts boolean or "true"/"false"; see below
}

Notes:

  • Do not include apiKey or url here; they resolve from your provider configuration.
  • Embedding models: If name includes "embed" or you pass "extractEmbeddingModels": true, the model is stored as an embedding model.
  • Requests using unsupported embedding models return 400 with a specific message (e.g., text-embedding-3-large is not supported).
  • programmaticToolCalling accepts a boolean or the strings "true"/"false"; an invalid value returns 400.

Programmatic tool calling ​

Programmatic tool calling (PTC) is Anthropic's "code execution"–style tool use, where the model writes and runs code to invoke tools instead of emitting normal tool-use blocks. In MissionSquad it is a property of the model configuration, set with the programmaticToolCalling field on POST /v1/core/add/model.

  • It is not a chat-completions request parameter and not an agent-config field. An agent inherits PTC from its underlying model.
  • When you omit programmaticToolCalling, it defaults based on the model: it is enabled automatically for Anthropic Claude Sonnet/Opus 4.6 and newer, and disabled for Haiku and for Claude 4.5 and earlier. Pass programmaticToolCalling: false to force it off, or true to force it on.
  • PTC requires tools to be available to the model and is enforced/mapped downstream by the provider adapter; it has no effect on providers other than Anthropic.

POST /v1/core/delete/model — Delete a model or embedding model ​

Body:

json
{ "modelId": "my-gpt4" }

See also ​