Ruby Q — the EVE Online character this project is named after

Open-source MCP server · MIT · 16 tools

EVE Mentor

Named for Ruby Q — born 2011, 0 kills, 8 losses, finally learning to fly.

An AI mentor for New Eden's learning cliff. It connects Claude to live EVE Online data — your real losses, your real skills, the real market — so you can finally just ask the game how to play it.

Get it on GitHub →

Why I built this

EVE Online is the deepest sandbox in gaming and the hardest one to learn. Players say it best: "In EVE, you don't have a learning curve, you have a cliff." Of every 10,000 new pilots, fewer than 500 are still logging in a month later. The knowledge exists, but it lives in wikis, veteran Discord servers, and a dozen third-party tools that all assume you already understand the game.

I created my character in August 2011. My lifetime record across fifteen years is 0 kills and 8 losses. I would resubscribe about once a year, play for a month, hit a wall that nobody explained, and quit. The walls were always the same. In 2015 I lost two cruisers in one day, the second one in Uedama, which I later learned is the most notorious gank chokepoint in the game (every veteran knows this; no tutorial mentions it). In 2018 I lost a destroyer that I had fitted with railguns, artillery, and blasters at the same time, three weapon systems from three different races on one hull. Nobody told me why that couldn't work. One year I came back, couldn't dock where I had left all my stuff, got no answer I understood, and cancelled again.

So I built the mentor I needed. On its first real session it pulled my losses from zKillboard and explained each death, told me the exploration frigate I always wanted was 16,000 skillpoints away (about ten hours of training), and found my lost property sitting in two asset-safety wraps that had been waiting for me for years. The docking mystery turned out to be a player-structure lockout, which the game also never explains.

That first session revealed one more thing: across all those failed attempts, past-me had quietly accumulated 17 million skillpoints and 702 million ISK. It turned out I was a mid-tier veteran who could never find the door, not a failed newbie. I suspect EVE is full of players like me, and this tool is for all of us.

What it does

It ships as a Model Context Protocol server, so it works with Claude Desktop, Claude Code, and any MCP-capable AI client. Built on the official ESI API, EVE SSO, and zKillboard. Free and open source (MIT).

Install it (5 minutes, no coding)

You need the Claude Desktop app and Node.js (free — download the LTS version and click through the installer).

  1. Download & build: open Terminal (Mac: Cmd+Space, type "Terminal") or PowerShell (Windows) and paste:
    git clone https://github.com/henryjrobinson/eve-mentor-mcp
    cd eve-mentor-mcp
    npm install && npm run build
  2. Tell Claude about it: edit claude_desktop_config.json (Mac: ~/Library/Application Support/Claude/ · Windows: %APPDATA%\Claude\) and add:
    { "mcpServers": { "eve-mentor": {
        "command": "node",
        "args": ["/FULL/PATH/TO/eve-mentor-mcp/dist/index.js"] } } }
    Then fully quit and reopen Claude.
  3. Test: ask Claude — "Pull the last 3 losses for character <your pilot> and explain what went wrong." No EVE login needed for that; the full step-by-step (including linking your character) is in the README.

Works with Claude Desktop, Claude Code, Cursor, LM Studio, and any MCP client. ChatGPT supports only remote MCP servers — a hosted version of the public tools is on the roadmap.

Roadmap

  1. v0.1 · ✅ Shipped

    Loss mentor

    Real losses from zKillboard with full fit detail, system danger intel, Jita prices, EVE SSO.
  2. v0.2 · ✅ Shipped

    "Can I fly it?" + career sorting hat

    Recursive skill-prerequisite plans diffed against your character, 15-path career matching, stranded-asset finder.
  3. v0.3 · ✅ Shipped

    Fitting & combat coach

    Mechanical fit checking, ammo damage-type matching per faction, slang glossary, session sitrep, live PLEX economics.
  4. v0.4 · ○ Planned

    Direction & ISK

    "What should I do tonight?" matched to your skills, ship, location, and risk appetite.
  5. v0.5 · ○ Planned

    Safety, logistics & news

    Route danger scoring, where to buy sensibly, and a "what happened while you were gone" briefing.

Follow along

Built by Henry Robinson. Questions, ideas, or want to test-fly it? Get in touch.