API Reference

The Unimatrix REST API — store and retrieve memory context across all LLMs and devices.

MCP Reference →MCP Quickstart →Security & Architecture →

Authentication

Every request must include your Unimatrix API key in the Authorization header. Keys are generated in Settings → API Keys after signing in.

curl https://www.deployunimatrix.com/api/spaces \
  -H "Authorization: Bearer umx_your_api_key_here"

Endpoints

Base URL: https://www.deployunimatrix.com/api · Raw OpenAPI spec: /api/openapi.json

GEThttps://www.deployunimatrix.com/api/spaces

List memory workspaces

Returns all memory workspaces owned by the authenticated user.

Responses

200List of spaces
401Unauthorized

POSThttps://www.deployunimatrix.com/api/spaces

Create a memory workspace

Request Body (required)

{
  "type": "object",
  "required": [
    "name"
  ],
  "properties": {
    "name": {
      "type": "string"
    },
    "description": {
      "type": "string"
    },
    "isPublic": {
      "type": "boolean"
    }
  }
}

Responses

201Space created
400Invalid input
401Unauthorized

GEThttps://www.deployunimatrix.com/api/spaces/{spaceId}

Get a space with its locations and memories

Path & Query Parameters

Name	In	Schema
spaceId *	path	string

Responses

200Space detail
404Not found

PATCHhttps://www.deployunimatrix.com/api/spaces/{spaceId}

Update a space

Path & Query Parameters

Name	In	Schema
spaceId *	path	string

Request Body (optional)

{
  "type": "object",
  "properties": {
    "name": {
      "type": "string"
    },
    "description": {
      "type": "string"
    },
    "isPublic": {
      "type": "boolean"
    }
  }
}

Responses

200Updated space
404Not found

DELETEhttps://www.deployunimatrix.com/api/spaces/{spaceId}

Delete a space

Path & Query Parameters

Name	In	Schema
spaceId *	path	string

Responses

200Deleted
404Not found

GEThttps://www.deployunimatrix.com/api/memories

List memories

Path & Query Parameters

Name	In	Schema
spaceId	query	string

Responses

200List of memories

POSThttps://www.deployunimatrix.com/api/memories

Store a new memory

Request Body (required)

{
  "type": "object",
  "required": [
    "spaceId",
    "content"
  ],
  "properties": {
    "spaceId": {
      "type": "string"
    },
    "content": {
      "type": "string"
    },
    "tags": {
      "type": "array",
      "items": {
        "type": "string"
      }
    }
  }
}

Responses

201Memory created

GEThttps://www.deployunimatrix.com/api/search

Full-text search across memories

Path & Query Parameters

Name	In	Schema
q *	query	string
spaceId	query	string

Responses

200Search results

POSThttps://www.deployunimatrix.com/api/locations

Create a location (room) inside a space

Request Body (required)

{
  "type": "object",
  "required": [
    "spaceId",
    "name"
  ],
  "properties": {
    "spaceId": {
      "type": "string"
    },
    "parentId": {
      "type": "string",
      "nullable": true
    },
    "name": {
      "type": "string"
    },
    "description": {
      "type": "string"
    },
    "position": {
      "type": "integer"
    }
  }
}

Responses

201Location created

GEThttps://www.deployunimatrix.com/api/tools

Discover available Unimatrix tools

Returns all memory/context tools in OpenAI function calling format by default. Add ?format=mcp to receive the raw MCP tool definitions instead. This endpoint is public (no auth required) because tool schemas are not sensitive.

Path & Query Parameters

Name	In	Schema
format	query	string · openai \| mcp · default: openai

Responses

200List of tools
500Failed to load tools from MCP backend

POSThttps://www.deployunimatrix.com/api/tools/call

Execute a Unimatrix tool (MCP REST fallback)

Translates a simple {toolName, args} payload into an internal MCP tools/call invocation. Use the exact tool names returned by GET /api/tools. Requires a valid Unimatrix API key. FOR NON-MCP LLMs: ALWAYS pass top-level 'sourceLlm' (e.g. 'gemini', 'chatgpt') in the body root so conversations auto-file into the per-LLM history location auto-created when you connected the provider during onboarding. The store_memory tool description (from /api/tools) contains the full system-prompt text to paste into your LLM.

Request Body (required)

{
  "$ref": "#/components/schemas/ToolCallRequest"
}

Responses

200Tool executed successfully
400Bad request (missing toolName or malformed args)
401Unauthorized (invalid or missing API key)
404Unknown tool
500Internal tool execution error

Universal Tool Endpoint

GET /api/tools and POST /api/tools/call translate between the REST API and the MCP tool surface for LLMs that support function calling but not MCP. Call GET /api/tools to discover schemas, then execute tools by name.

For non-MCP LLMs, always pass sourceLlm (e.g. "gemini", "chatgpt") at the top level of the /api/tools/call body so stored memories auto-file into the correct per-LLM history location.

Learn how MCP clients connect →