eng-leadership-toolkit

"I just became an engineering manager. What should I focus on?""What's a healthy sprint completion rate?""Do I need a mentor or a coach?""What is this developer worth on the market?"

Your AI already gets asked these. Now it can answer with first-party data from 3,400+ paid 1:1 mentoring sessions with 300+ engineering leaders in 17+ countries, instead of guessing.

This is a remote MCP server. No install, no API key. One URL:

https://www.marian.coach/mcp

Tools

Tool	The question it answers
`calculate_developer_value`	What is this developer worth? Scores 15 skills across 5 pillars, returns level + 2026 Western-Europe salary estimate
`get_engineering_leadership_benchmarks`	What's healthy? Sprint completion, roadmap %, manager time per report, mentee mix, 2025 topic demand. CC BY 4.0, citable
`choose_mentor_coach_or_advisor`	Mentor, coach, or advisor? All three compared + a three-question self-test
`get_one_on_one_playbook`	How do I run this 1:1? Real session templates: first session, underperformance, promotion to EM, better 1:1s, career move
`get_first_time_manager_guidance`	I'm a new EM, now what? The responsibility triangle, six failure modes, readiness self-check

Connect

Claude Code

claude mcp add -t http eng-leadership-toolkit https://www.marian.coach/mcp

Claude.ai / Claude Desktop

Settings > Connectors > Add custom connector > https://www.marian.coach/mcp

Cursor (.cursor/mcp.json)

{
  "mcpServers": {
    "eng-leadership-toolkit": {
      "url": "https://www.marian.coach/mcp"
    }
  }
}

ChatGPT

Settings > Connectors > Developer mode > Add server > https://www.marian.coach/mcp

Human-readable docs live at the same URL: open www.marian.coach/mcp in a browser.

Where the data comes from

Every response is grounded in Marian Kamenistak's mentoring practice, not model guesses:

The calculator mirrors the live developer value calculator
Benchmarks come from the published engineering leadership statistics, licensed CC BY 4.0
Playbooks are the actual session templates used across 3,400+ sessions

Nothing is synthetic. Thin topics were cut rather than padded.

Stack

Cloudflare Workers + Durable Objects, McpAgent from the agents SDK, streamable HTTP, authless. GET /mcp with an HTML accept header serves the docs page; MCP clients hit the same URL.

npm install
npx wrangler dev      # local
npx wrangler deploy   # ship it

Author

Marian Kamenistak. 20+ years in software, grew Mews engineering 8 to 80 through Series C, now mentors engineering leaders full-time and runs the Engineering Leaders Community in CEE.

Found it useful? A star helps other engineering leaders' AIs find it too.

License

MIT

eng-leadership-toolkit