Skillget

biomed-agent

mcp
Security Audit
Pass
Health Pass
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 15 GitHub stars
Code Pass
  • Code scan — Scanned 11 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
This tool acts as an orchestrator and registry, connecting AI agents (like Claude Code) to various external biomedical research data sources such as OpenTargets, Monarch, and MyGene.

Security Assessment
Overall Risk: Low. The light code scan across 11 files found no dangerous patterns, hardcoded secrets, or dangerous permission requests. The architecture relies on making network requests to external public APIs to fetch biomedical data. It does not appear to execute arbitrary shell commands or interact with local sensitive files. However, the setup requires configuring external sibling repositories via local paths, so standard caution should be applied when installing those dependencies.

Quality Assessment
Quality: Good. The project is actively maintained (last updated today) and is protected by the permissive Apache-2.0 license. It has earned 15 GitHub stars, indicating a small but real level of community trust. The repository is well-documented with clear usage instructions, test suites, and transparent safety disclaimers explicitly stating it is for research and not clinical decision-making.

Verdict
Safe to use for educational and biomedical research purposes.
SUMMARY

Connecting AI agent to biomedical data

README.md

Biomedical Agent Workspace

Agent-facing biomedical research workspace for Codex and Claude Code.

This repo no longer owns an LLM reasoning loop or Streamlit UI. Codex or Claude Code should call the five biomedical MCP servers directly, guided by AGENTS.md and skills/biomed-research/SKILL.md.

What Stays Here

  • MCP server registry and local path handling.
  • A tiny diagnostics CLI for server/tool inspection.
  • The cross-server research contract for agents.
  • Example MCP registration in mcp.json.

Architecture

flowchart TB
    user["User"] --> runtime["Codex / Claude Code"]

    subgraph repo["biomed-agent"]
        contract["AGENTS.md<br/>canonical contract"]
        skill["skills/biomed-research<br/>research workflow"]
        config["mcp.json<br/>server registration"]
        diagnostics["Diagnostics CLI<br/>list, inspect, call, doctor"]
        client["core/servers.py + core/mcp_client.py<br/>mechanical MCP access"]
    end

    runtime --> contract
    contract --> skill
    contract --> config
    diagnostics --> client

    config --> servers["5 biomedical MCP servers"]
    client --> servers
    servers --> sources["OpenTargets, Monarch, MyGene,<br/>MyChem, MyDisease data"]

MCP Servers

The default setup expects sibling repos:

  • ../opentargets-mcp
  • ../monarch-mcp
  • ../mygene-mcp
  • ../mychem-mcp
  • ../mydisease-mcp

Override paths with OPENTARGETS_MCP_PATH, MONARCH_MCP_PATH, MYGENE_MCP_PATH, MYCHEM_MCP_PATH, or MYDISEASE_MCP_PATH.

Setup

uv sync

Diagnostics

uv run python -m ui.cli list-servers
uv run python -m ui.cli list-tools
uv run python -m ui.cli list-tools --server opentargets
uv run python -m ui.cli call-tool opentargets.search_entities '{"query_string":"BRAF","entity_names":["target"]}'
uv run python -m ui.cli doctor

Tests

uv run pytest tests/ -q

Safety

Research and educational use only. This is not a clinical decision system and must not provide diagnosis, prescribing, dosing, or patient-specific treatment advice.

Reviews (0)

No results found