Jev MCP server

The Jev MCP server drops a typed decision tool into any Model Context Protocol client. Instead of the client guessing and parsing, it calls one tool and gets back a locked answer — one of your options, calibrated probabilities, and confidence.

MCP lets an agent call external tools; Jev is the tool you want for any yes/no, pick-one, or classify decision. Jev is TypeSafe AI's System One model, so the tool returns a typed result — never free text the client has to re-interpret — in roughly 70–500ms for about $0.001 a call.

Run it

// Run the Jev MCP server straight from GitHub — no install step:
npx github:codaaiteam/jev-mcp

// Set your key in the environment first:
export JEV_API_KEY=jv_live_...

Point your MCP client (Claude Code, Claude Desktop, or any MCP host) at that command. The server registers a decide tool the model can call whenever it needs a typed, calibrated answer instead of a guess.

Why an MCP decision tool

• The client gets one option locked to your list — no invalid labels
• Calibrated confidence (RLCD) the agent can gate on
• One round trip, ~$0.001, no parsing on the client side
• Runs over npx from github: — nothing to build or host

Under the hood

The MCP server is a thin wrapper over the same endpoint everything else uses: POST https://jevtypesafeai.com/api/v1/decide with a jv_live_ key (official upstream POST https://api.typesafe.ai/v1/systemone). See the full integration guide for config options and tool schemas.

See also: MCP integration guide · Jev for AI agents · Docs