Connect myai to Your Agent via MCP

myai exposes its context — artifacts, dimensions, and work orders — over the Model Context Protocol (MCP). Plug the endpoint into the agent of your choice and that agent can read from your graph in real time.

Endpoint

Your instance exposes an MCP endpoint at:

https://api-[instance].makeyourself.ai/api/mcp

Replace [instance] with your tenant subdomain (the same one you use to reach the myai web app).

What's exposed

The MCP surface gives your agent:

• Artifacts — typed nodes in your graph, with relationships and lineage
• Dimensions — scoped lenses that determine which artifacts and skills are in context
• Work orders — durable session records that survive context-window compaction

The agent reads through MCP using the standard tool surface — no special myai SDK needed on the agent side.

Plug it into Claude

In Claude (Desktop or claude.ai with MCP enabled), add a new MCP server:

{
  "mcpServers": {
    "myai": {
      "url": "https://api-[instance].makeyourself.ai/api/mcp"
    }
  }
}

Plug it into ChatGPT

In ChatGPT's MCP settings, add a new connector pointing at:

https://api-[instance].makeyourself.ai/api/mcp

Plug it into Cursor

In Cursor's MCP settings (Settings → MCP → Add server), add:

{
  "myai": {
    "url": "https://api-[instance].makeyourself.ai/api/mcp"
  }
}

What you get

Once connected, your agent can:

• Search artifacts by name, content, or relationships
• Pull a specific artifact's body and lineage
• Read what's in scope for a given dimension
• Reference work orders so context survives across sessions

For more on what's actually in the graph that you'd be exposing, see the platform overview.

Why this exists

We don't try to replicate the MCP integration ecosystems your agents already have. myai is the context; your agent of choice is the hands. Read more in the blog post: myai is the context. Your agent is the hands..