literature-agent-mcp

MCP server that exposes the local biomedical literature pipeline in:

• /Users/zouxingyu/Desktop/Codex工作区/literature-agent

It wraps the existing five-step CLI workflow as MCP tools so OpenClaw / Codex / Claude Code / other MCP clients can call it as a standard server.

Tools

• plan_query

• search_and_merge

• download_fulltexts

• convert_and_index

• write_draft

• run_pipeline

Default assumptions

• Biomedical / pathology domain

• Legal open-access retrieval only

• No paywall bypass

• Local literature-agent project already exists

Local setup

cd /tmp/mcp_repo_zouxy111
python3.11 -m venv .venv
source .venv/bin/activate
pip install -e .

Run server

source .venv/bin/activate
literature-agent-mcp

Environment variables

• LITERATURE_AGENT_ROOT: override the default project root
  Default: /Users/zouxingyu/Desktop/Codex工作区/literature-agent

• LITERATURE_AGENT_PYTHON: override the interpreter used to call the backend
  Default: <root>/.venv/bin/python, fallback python

Example MCP config

{
  "mcpServers": {
    "literature-agent": {
      "command": "/tmp/mcp_repo_zouxy111/.venv/bin/literature-agent-mcp",
      "env": {
        "LITERATURE_AGENT_ROOT": "/Users/zouxingyu/Desktop/Codex工作区/literature-agent"
      }
    }
  }
}

Backend outputs

For a given task_id, outputs are written under:

<literature-agent>/work/<task_id>/

Typical structure:

• query_plan.json

• search/master_results.jsonl

• corpus/

• draft/