MyPlayground

MCP Server for querying compensation data from levels.fyi.

What this server provides

Server name: LevelsFyi

Tools:

• get_recent_offers(company_name: str, role: str, level: str, location: str = None) -> dict (Hit levels.fyi salary search API to find the recent offer data for given company, role, level, location)
• get_level_mapping(company_name: str, role: str = "Software Engineer") -> dict (Get the level mapping for a given job family at a company)

End User Installation

This section is for users who only want to install and run the MCP server in Cursor.

Requirements

• Python >=3.11
• uv

Installing the MCP

This MCP is hosted on Prefect Horizon, which is a low-code/no-code way to host MCP servers for LLMs to be able to use. The MCP URL is https://levels-fyi.fastmcp.app/mcp, so you should be able to use this server address to add to Codex, Claude Code, OpenAI SDK, Cursor, Gemini CLI.

For example, in Claude Code, you can reference this wiki for how to install the MCP server. You can run claude mcp add --transport http levels-mcp https://levels-fyi.fastmcp.app/mcp to install the MCP server.

If you're on Cursor, you can visit this link and follow the on-screen instructions

Developer Setup

This section is for contributors working on code changes.

Install dependencies

From the project root:

uv sync

Run server locally

uv run python main.py

Add or update dependencies

uv add <package>
uv sync

Quick environment check

uv run python -c "import fastmcp; print('fastmcp ok')"

Project structure

.
├── main.py
├── src/
├── pyproject.toml
└── README.md

Troubleshooting

Failed to spawn: fastmcp (os error 2)

Cause: Cursor cannot find the fastmcp executable.

Fix:

1. Prefer Cursor config with uv run python main.py, or
2. Install global tool with uv tool install fastmcp and ensure PATH includes the uv tool bin (commonly ~/.local/bin).

Contributing

• If you have a feature request, please create a GitHub issue to triage the requests
• Feel free to fix any issues but please raise a Pull Request so that I can review it prior to deployment