jita-mcp

An MCP server that lets Claude (and any other MCP-compatible LLM) actually fit EVE Online ships. The LLM brings the meta knowledge — what's good for what content — and jita-mcp answers the ground-truth questions: does this fit work, what does it actually do, what does it cost?

Stats are computed by pyfa's fitting engine, so ship bonuses, skill multipliers, ammo selection, drone DPS, implant effects, and stacking penalties are all applied correctly. Module prices come live from ESI (Jita 4-4 sell-min, cached per ESI's rules).

What you can ask Claude

"Build me a kinetic-DPS Condor at All V. Use jita-mcp."

"I have a Vexor with Hammerhead IIs. What's the best low-slot DPS mod and tank for a buffer fit? Keep it under 50M ISK."

"Validate this Manticore fit and tell me if it's cap-stable."

"Export this fit to EFT and list every skill I need."

Related MCP server: EVE ESI Tool

Tools

The full schemas are exposed via MCP's list_tools; Claude reads them automatically.

Quick start

Requires Python 3.12+ and uv.

# 1. Clone with submodules (pyfa is vendored)
git clone --recurse-submodules https://github.com/zachyt/jita-mcp
cd jita-mcp

# 2. Install deps + build pyfa's data DB (one-time, ~10s)
uv sync
uv run python scripts/build_eve_db.py

# 3. (Optional) Download the Fuzzwork SDE if you want it ready for future
#    universe-data tools
python3 scripts/fetch_sde.py

# 4. Start the server
uv run jita-mcp                              # listens on :8080

Adding it to Claude Code

claude mcp add jita-mcp --transport http http://localhost:8080/mcp

Restart your Claude Code session and the five tools become available. After that, just talk to Claude about EVE fits — it'll call the tools as needed.

Adding it to Claude.ai (web)

Settings → Connectors → Add custom connector → point at your server's HTTPS URL. Requires public HTTPS, so you'll want to deploy it (see Deployment below) or use a tunnel like ngrok.

Day-to-day commands

Configuration

All via environment variables (defaults in jita_mcp/config.py):

Deployment

Build the production container with:

docker build -t jita-mcp:latest .
docker run -p 8080:8080 jita-mcp:latest

The Dockerfile is multi-stage and bakes everything needed (eve.db, the eos source, our Python deps) into the runtime image. Approximate sizes: ~400 MB compressed, ~2 GB on disk.

Point any HTTPS-fronted host (Fly.io, Lightsail, your own VPS) at the container and add the URL as a custom connector in Claude.

License

GPL-3.0-or-later. The project links pyfa's eos engine (LGPL-3.0) and ships pyfa's bundled data; GPL-3.0 satisfies both.

Credits

• Pyfa — the fitting engine and the EVE data pipeline this server depends on

• Fuzzwork — the SDE database dumps

• CCP — for keeping ESI public and free