Connect an agent

Give an AI tool access to your GigaRAG Workspace, so it can search, read and write your memos while it works.

OpenAI API

Connect GigaRAG as a remote MCP tool in a Responses API request.

Use it in code

The snippet has your key in it so it runs as pasted. Move it to an environment variable before the code is committed or shared.

Send a request with GigaRAG as an MCP tool

OpenAI does not store the header, so send it with every request.

curl https://api.openai.com/v1/responses \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5",
    "input": "List my GigaRAG buckets.",
    "tools": [{
      "type": "mcp",
      "server_label": "gigarag",
      "server_url": "https://mcp.gigarag.com/mcp",
      "headers": { "Authorization": "Bearer <api-key>" },
      "require_approval": "never"
    }]
  }'

Check it worked

Send this once it is connected. An answer drawn from your workspace means the connection is live.

Run it. The model should answer with your GigaRAG buckets.

Written from the official docs: Connectors and MCP