Lacuna Research MCP

Give your coding agent a map of machine learning research.

Search papers, trace Research Directions, inspect research proposals, and follow every derived object back to its source in Lacuna.

Install the server View the source

Works with Codex, Claude Code, Claude Desktop, and other stdio MCP clients.

research session

$ codex mcp add lacuna-research   -- uvx --from git+https://github.com/tiptreesystems/lacuna-research-mcp.git   lacuna-research-mcp

# Then ask your agent:
Trace the Research Direction on autonomous
LLM agents for scientific discovery.

Install

One server, one command.

The commands below use uvx to run the server directly from GitHub. You need uv and Python 3.11 or newer.

Codex

codex mcp add lacuna-research -- uvx --from git+https://github.com/tiptreesystems/lacuna-research-mcp.git lacuna-research-mcp

Run codex mcp list to verify the connection.

Claude Code

claude mcp add --scope user lacuna-research -- uvx --from git+https://github.com/tiptreesystems/lacuna-research-mcp.git lacuna-research-mcp

Run claude mcp get lacuna-research to verify the connection.

Configuration for other MCP clients

{
  "mcpServers": {
    "lacuna-research": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/tiptreesystems/lacuna-research-mcp.git",
        "lacuna-research-mcp"
      ]
    }
  }
}

Or ask your agent

Install and configure the Lacuna Research MCP from https://github.com/tiptreesystems/lacuna-research-mcp for this client.

What it exposes

Research objects your agent can work with.

Search the map

Search papers, Research Directions, proposals, authors, venues, and institutions.

Move through a field

Open a Research Direction, inspect its papers, and follow neighboring directions.

Read the evidence

Pull compact paper context, concepts, figures, and links back to the source.

Find the people

Inspect an author's papers, current directions, research neighbors, and institutions.

A useful first task

Start from a connection, then check the papers.

Lacuna groups papers through their concept elements. That lets an agent find a shared scientific move across papers that use different language or sit in different subfields.

Trace the Research Direction “Autonomous LLM Agents for Scientific Discovery.” Compare the chemistry robotics, physics discovery, and alloy design papers. Cite the Lacuna page and source paper for each claim.

How it works

A thin adapter over Lacuna's public API.

The open-source server normalizes Lacuna IDs and URLs, requests bounded agent-oriented context, retries transient failures, and returns structured data with source links. The agent decides which search and detail tools to call as the research task develops.

search_lacuna get_direction get_direction_papers get_paper get_author_context get_hypothesis

GitHub repository Tool reference Lacuna paper Alien Science