MissionSquad API Overview

The MissionSquad API is OpenAI-compatible for Chat Completions and Embeddings while using API keys via the x-api-key header. All resources (models, agents, embedding models, vector stores, files) are namespaced and isolated per account.

• Base URL: set to your deployment API origin (examples use https://agents.missionsquad.ai/v1)
• Authentication: x-api-key: <your-msq-api-key>
• OpenAI SDK compatibility: Authorization: Bearer msq-... is accepted for API keys starting with msq- (see below)

Authentication

Include your MissionSquad API key with every request.

• Required header: x-api-key: msq-****************************
• Content types:
  • JSON: Content-Type: application/json
  • Uploads: Content-Type: multipart/form-data

Examples:

curl -H "x-api-key: $MSQ_API_KEY" https://agents.missionsquad.ai/v1/models

// Node 20+ fetch
const res = await fetch("https://agents.missionsquad.ai/v1/models", {
  headers: { "x-api-key": process.env.MSQ_API_KEY! }
});

OpenAI SDK (compatibility path)

When using the official openai package, configure baseURL to your MissionSquad API and pass your MissionSquad API key as the SDK's apiKey. The SDK will send Authorization: Bearer msq-..., which MissionSquad accepts for API keys that start with msq-.

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.MSQ_API_KEY, // Authorization: Bearer msq-...
  baseURL: "https://agents.missionsquad.ai/v1"
});

const completion = await client.chat.completions.create({
  model: "my-gpt4", // model, your own published agent name, or shared/<ownerUsername>/<slug>
  messages: [
    { role: "system", content: "You are helpful." },
    { role: "user", content: "Hello!" }
  ]
});

console.log(completion.choices[0].message);

Namespacing and Isolation

• All user-defined entities are scoped to your account: models, agents, embedding models, vector stores, files.
• List/CRUD operations only access resources associated with your API key.
• Chat resolves your own models and published agents in your namespace, and can also target explicitly shared published agents via shared/<ownerUsername>/<slug>.

OpenAI-Compatible Endpoints

• POST /v1/chat/completions
• POST /v1/chat/cancel
• POST /v1/embeddings

Core audio endpoints:

• POST /v1/core/audio/tts
• POST /v1/core/audio/tts/stream
• POST /v1/core/audio/stt

Works with:

• Official OpenAI SDK (JavaScript/TypeScript)
• Any OpenAI-like API client (set baseURL to your MissionSquad API and supply your key)

See detailed reference pages in the sidebar.

Error Handling Conventions

• 400 — Missing/invalid parameters (e.g., missing model, invalid name, unsupported embedding model)
• 401 — Unauthorized (missing/invalid API key)
• 404 — Resource not found (model, agent, vector store, or file does not exist)
• 409 — Conflict (e.g., adding a file already associated with the vector store)
• 500 — Internal errors

Provider API keys and credentials are masked in all read APIs.

Patterns and Tips

• Prefer x-api-key for direct API usage.
• For OpenAI SDK usage, provide your MissionSquad API key as apiKey with baseURL set to your API origin (Authorization: Bearer msq-... supported).
• Streaming chat uses SSE with OpenAI-like chunk payloads (text/event-stream). In Node 20+, use ReadableStreamDefaultReader via res.body.getReader(). In browsers, consider EventSource or streaming-compatible readers.
• Chat completions support structured output hints via response_format/responseFormat and provider passthrough values via extra_params/extraParams.
• Audio endpoints support TTS (binary and SSE streaming) plus STT multipart uploads.
• Unsupported embeddings: text-embedding-3-large is not supported. Use text-embedding-3-small or nomic-embed-text-v1.5. Requests with unsupported models will return 400 with a message.

Endpoint Index

Core:

• GET /v1/models
• GET /v1/modelmap
• POST /v1/chat/completions
• POST /v1/embeddings
• POST /v1/core/audio/tts
• POST /v1/core/audio/tts/stream
• POST /v1/core/audio/stt
• GET /v1/core/providers
• POST /v1/core/add/provider
• POST /v1/core/delete/provider
• POST /v1/core/models
• POST /v1/core/add/model
• POST /v1/core/delete/model
• GET /v1/core/agents
• POST /v1/core/add/agent
• POST /v1/core/delete/agent
• POST /v1/core/agents/publish
• GET /v1/core/agents/published
• GET /v1/core/agents/shared-with-me
• POST /v1/core/agents/:username/:slug/shares
• GET /v1/core/agents/:username/:slug/shares
• DELETE /v1/core/agents/:username/:slug/shares
• POST /v1/core/generate/prompt
• POST /v1/core/agent-workflow
• GET /v1/core/workflows
• POST /v1/core/workflows
• PUT /v1/core/workflows/:id
• DELETE /v1/core/workflows/:id
• POST /v1/core/workflow-runs
• GET /v1/core/workflows/:id/runs
• GET /v1/core/workflow-runs/:runId
• GET /v1/core/workflow-runs/:runId/hydrated
• GET /v1/core/workflow-runs/:runId/stream
• POST /v1/core/workflow-runs/:runId/cancel
• GET /v1/core/config
• POST /v1/core/scrape-url
• GET /v1/core/tools
• GET /v1/core/servers
• GET /v1/core/usage/raw
• GET /v1/core/usage/summary
• GET /v1/core/usage/models
• GET /v1/core/usage/timeseries
• GET /v1/core/usage/billing
• POST /v1/core/collections/:collectionName/search
• GET /v1/core/collections/:collectionName/diagnostics
• POST /v1/core/collections/:collectionName/recover
• GET /v1/core/collections

Vector Stores & Files:

• GET /v1/vector_stores
• POST /v1/vector_stores
• GET /v1/vector_stores/:id
• DELETE /v1/vector_stores/:id
• GET /v1/vector_stores/:id/files
• POST /v1/vector_stores/:id/files
• GET /v1/vector_stores/:id/files/:fileId
• POST /v1/vector_stores/cancel
• GET /v1/files
• POST /v1/files (multipart)
• GET /v1/files/:id
• DELETE /v1/files/:id
• GET /v1/files/:id/content
• GET /v1/user-collections
• GET /v1/vector_stores/:id/file-details

Published/shared agent access:

• GET /v1/public/agent/:username/:slug
• POST /v1/public/agent/:username/:slug/chat
• POST /v1/public/agent/:username/:slug/speak
• GET /v1/public/agent/:username/:slug/sessions
• GET /v1/public/agent/:username/:slug/sessions/:id
• POST /v1/public/agent/:username/:slug/sessions
• PATCH /v1/public/agent/:username/:slug/sessions/:id/title
• DELETE /v1/public/agent/:username/:slug/sessions/:id

Notes:

• The /v1/public/agent/... routes are authenticated in the current API. They are accessible to the owner or a user/email recipient that has been explicitly shared on the agent.
• See Agents for the verified request and response shapes for publishing, sharing, public invocation, and public session persistence.

MCP API (Admin)

The MCP API is a separate, admin-only companion service deployed as a sidecar container alongside the MissionSquad API. It is not publicly exposed and is intended for instance operators.

The MissionSquad API communicates with the MCP API internally via the TOOLS_HOST environment variable. The MCP API manages:

• MCP server package installation and lifecycle (Node.js, Python, Streamable HTTP)
• Server registration, configuration, enable/disable
• Encrypted per-user secret storage and automatic injection into tool calls
• OAuth authentication flows for tool servers

See the full MCP API documentation for endpoint reference and deployment details.

See also

• Chat Completions
• Audio (TTS/STT)
• Embeddings
• Providers
• Models
• Agents
• Usage
• Core Utilities
• Collections
• Vector Stores
• Files
• Webhooks
• Convenience
• Endpoint Index
• MCP API (Admin)