typesafe-mcp
Go CLI and single-binary MCP server exposing TypeSafe judgments to Claude Desktop, Claude Code, and Codex.
by @0xLogicrw
What it does
typesafe-mcp in the Jev ecosystem. Go CLI and single-binary MCP server exposing TypeSafe judgments to Claude Desktop, Claude Code, and Codex. Agent plumbing where probabilities beat regex on model output. Fork the pattern, not the prose.
Tags in the wild: agent tooling, mcp, agents. The repo has install steps, env vars, and recent commits if you are deciding whether to depend on it.
How Jev is used
The evaluate binary exposes System One as MCP tools so agents return probabilities they can branch on. evaluate setup mcp registers hosts it detects; evaluate setup pi covers the Pi agent.
Setup: curl install.sh to ~/.local/bin, export PATH, set TYPESAFE_API_KEY or gateway keys per README. Go install github.com/itsmostafa/typesafe-mcp/cmd/evaluate@latest also works.
Why agents want typed evaluate
Free-text tool outputs force brittle parsing. typesafe-mcp returns structured probabilities for Choice, Score, and Noul questions so your agent can threshold, log, and retry deterministically.
FAQ
- What is typesafe-mcp?
A Go CLI and MCP server (evaluate) that wraps TypeSafe Jev for Claude Code, Claude Desktop, Codex, and Pi. It is the listing people mean when they search typesafe-mcp.
- How is this different from other Jev MCP servers?
It focuses on a single-binary install and setup subcommands for multiple hosts. Compare tool schemas and transport (stdio vs HTTP) before you standardize.
- Which API key do I need?
Follow the repository README for TypeSafe direct vs gateway variables. Keys stay on your machine; the server forwards asks to System One.
- Can Pi use it without Claude?
Yes via evaluate setup pi when Pi is installed locally.
Related products
Hand-picked neighbors with rich profiles or overlapping tags.