- No API key. It talks to the same API the explorer uses.
- Local & lightweight. Runs on your machine over stdio via
npx— nothing to host. - Mainnet & Sepolia. Switch networks with one env var.
Tools & usage
The full tool catalog and how an agent should chain calls
Explorer
The Voyager block explorer
Prerequisites
- Node.js 20.3 or newer (
npxships with it). - An MCP-capable client: Claude Code, Claude Desktop, Cursor, or Codex.
voyager-mcp. npx -y voyager-mcp downloads
and runs it for you — there is nothing to install or build.
Setup
- Claude Code
- Claude Desktop
- Cursor
- Codex
Add the server with one command (run it in your project, or with Or commit it to your repo so the whole team gets it — create
-s user to make it global):.mcp.json in the project root:.mcp.json
Windows: if
npx isn’t found, use "command": "npx.cmd" (or wrap with
cmd /c npx -y voyager-mcp).Choose a network
The server defaults to mainnet. To target Sepolia testnet, setVOYAGER_NETWORK:
~/.codex/config.toml instead:
Try it
Once the server is connected, ask your agent things like:- “Use voyager to get the current Starknet network stats.”
- “Look up Starknet transaction
0x…and the events it emitted.” - “Show me the verified Cairo code for class
0x….” - “What’s the class hash and type of contract
0x…?” - “Who are the top holders of token
0x…?” - “Is token
0x…safe? Run a security scan.”
Verify the connection
- Claude Code: run
/mcp—voyagershould be listed as connected. - Claude Desktop / Cursor: the tools appear in the tools/MCP menu once the client restarts.
- Codex: run
codex mcp list—voyagershould be listed.
npx -y voyager-mcp runs from your
terminal (the first run downloads the package, which can take a few seconds).