Quickstart

Get Buildd Memory connected to your AI agents in under 5 minutes.

1. Get an API Key

Generate an API key at buildd.dev/memory. Your key will look like mem_xxx.

If you're self-hosting, see Self-Hosting to set up your own instance and create keys via the root API key.

2. Add the MCP Server

Add Memory to your agent's MCP configuration. This works with Claude Code, Cursor, and any MCP-compatible client.

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["@buildd/memory"],
      "env": {
        "MEMORY_API_URL": "https://memory.buildd.dev",
        "MEMORY_API_KEY": "mem_your_key_here"
      }
    }
  }
}

Cursor

Add the same configuration to your Cursor MCP settings.

Direct Execution

You can also run the MCP server directly:

MEMORY_API_URL=https://memory.buildd.dev MEMORY_API_KEY=mem_xxx npx @buildd/memory

HTTP Transport (Remote / Serverless)

If you can't run a stdio process (e.g. remote agents, web apps, or serverless environments), use the HTTP MCP endpoint directly:

POST https://memory.buildd.dev/api/mcp
Authorization: Bearer mem_your_key_here
Content-Type: application/json

Send standard MCP JSON-RPC requests in the body. The endpoint is stateless and serverless-compatible — no persistent process needed. You can also authenticate with the x-api-key header instead of Authorization: Bearer.

3. Save Your First Memory

Once connected, your agent has access to the memory tool. Try saving a memory:

Use the memory tool to save a gotcha:
- type: gotcha
- title: "API rate limits are per-key, not per-IP"
- content: "Rate limiting is applied per API key. Multiple agents sharing a key share the rate limit."

4. Load Context

At the start of any session, load existing memories:

Use the memory tool with action "context" to load relevant memories.

This returns all memories formatted as markdown, ready for your agent to use.

5. Search Memories

Find specific memories when needed:

Use the memory tool to search for memories about "rate limit"

Next Steps

• Memory Types — Learn when to use each memory type
• Agent Workflows — Best practices for agent memory usage
• API Reference — Use the REST API directly