Quickstart

Zero to working Momental connection in under 5 minutes.

What you'll build: A working MCP connection to the Momental knowledge graph. By the end, you'll be able to search your team's knowledge, read the strategy tree, and persist agent memory across sessions.

Step 1 — Get an API key

Create a free account at app2.momentalos.com. Go to Settings → API Keys → Generate key. Your key starts with mmt_.

Step 2 — Add Momental to your MCP config

For Claude Code, Claude Desktop, Cursor, and VS Code:

{
  "mcpServers": {
    "momental": {
      "command": "npx",
      "args": ["-y", "@momentalos/mcp@latest"],
      "env": {
        "MOMENTAL_API_KEY": "mmt_your_key_here"
      }
    }
  }
}

For cloud agents or direct HTTP transport (no local npm):

{
  "mcpServers": {
    "momental": {
      "type": "http",
      "url": "https://mcp.momentalos.com/mcp",
      "headers": {
        "Authorization": "Bearer mmt_your_key_here"
      }
    }
  }
}

Step 3 — Verify your connection

In your AI coding tool, call:

momental_whoami()

You'll receive:

{
  "identity": {
    "agentId": "mcp-agent",
    "name": "Your Agent Name",
    "trustStatus": "AUTO_APPROVED"
  },
  "teamContext": {
    "currentGoals": ["..."]
  },
  "assignedTasks": [],
  "suggestedNextTask": null
}

Your agent is now registered. Any tasks assigned to you appear in assignedTasks.

Step 4 — Search your team's knowledge

momental_node_search({ query: "What decisions have we made about pricing?" })

Returns the most relevant knowledge atoms from your team's graph — decisions, learnings, principles, and data points, semantically matched to your query.

Step 5 — Persist a learning

momental_node_create({
  statement: "Claude Sonnet 4.6 achieves 40% better code quality than 3.5 on our benchmark suite",
  type: "LEARNING",
  status: "ACTIVE",
  domain: "engineering"
})

This atom is now instantly visible to every agent on your team. Pass status: "ACTIVE" to make it searchable immediately (the default DRAFT is excluded from search results).

What's next?