Agent Memory MCP

Use Engra Agent Memory from MCP clients (Cursor, Claude Desktop, etc.) via Model Context Protocol Streamable HTTP. See the REST API guide for HTTP integration.

MCP endpoint

Single URL supporting POST / GET / DELETE (stateless JSON mode). Clients send JSON-RPC messages to this endpoint.

https://ai-sass-7pymefszf-keyskulls-projects.vercel.app/api/v1/memory/mcp

Authentication

MCP and REST share the same dashboard API keys (inf_ prefix). Tenant-scoped keys are required; local .env placeholder keys cannot be used for production MCP.

Create an API key

Open Console → API keys (/dashboard/keys)
Multi-select scopes under Gateway / Memory groups
Quick presets: MCP read-only, MCP read/write, Gateway default
Read-only agents: memory:read only
For memory_save_atom: add memory:write
Memory MCP usually needs no Gateway binding
Copy the inf_… key immediately into your MCP client Authorization header

Scope reference

memory:read — memory_list_libraries / search / recall / wake-up / list_atoms
memory:write — memory_save_atom, POST /atoms

Headers

Authorization: Bearer inf_xxxxxxxx (recommended)
X-API-Key: inf_xxxxxxxx
Do not use upstream Gemini/OpenAI keys — those belong in Gateway upstream settings

Transport

Protocol: MCP Streamable HTTP (2025-11-25)
Mode: stateless + JSON response (one POST → one JSON body, no long-lived SSE required)
Standard MCP methods supported: initialize, tools/list, tools/call, etc.

Recommended workflow

Call memory_list_libraries first to obtain memoryLibraryId
Use memory_search / memory_recall / memory_wake_up for retrieval
Use memory_save_atom for synchronous writes
All tools except list_libraries require memoryLibraryId in arguments

Tools

Tool	Scope	Description
memory_list_libraries	memory:read	List memory libraries for the workspace (id, name, slug, persona)
memory_search	memory:read	L3 deep semantic search
memory_recall	memory:read	L2 topic recall
memory_wake_up	memory:read	L0+L1 wake-up stack
memory_save_atom	memory:write	Save an atom synchronously (same as ?sync=1)
memory_list_atoms	memory:read	Paginated atom list

Cursor configuration example

{
  "mcpServers": {
    "engra-memory": {
      "url": "https://ai-sass-7pymefszf-keyskulls-projects.vercel.app/api/v1/memory/mcp",
      "headers": {
        "Authorization": "Bearer inf_xxxxxxxx"
      }
    }
  }
}

HTTP integration

Custom MCP clients should follow the Streamable HTTP spec: POST requests need Accept including application/json and text/event-stream. This server primarily responds with application/json.

← Agent Memory REST API · API keys · Memory console