Jev and MCP: using Jev as a Model Context Protocol tool

Is there an official TypeSafe MCP server?

No, not as of September 2026. The docs index lists the HTTP API, the Python and JavaScript SDKs, and an agent skill for Claude Code and Codex, but no MCP server (docs index). TypeSafe's public GitHub organization lists the skill, the Python and JavaScript SDKs and an adapter library, alongside a few unrelated repos, and none of them is an MCP server (typesafe-ai on GitHub).

So every "Jev MCP" or "TypeSafe MCP" server you find is a community project. Read its README, check which API key it expects, and treat it as third-party code. For the official ways in, see how to use Jev and the SDKs and integrations list.

What a Jev MCP server actually does

Jev is a decision model, not a chat model. You send it a state (text, a JSON object or an array of text) plus typed questions, and it returns a typed answer for each one: a choice from your options, a score on your rubric, or a noul between 0 and 1 for a yes/no statement (Introduction). It does not write text, call tools or hold a conversation (Jev with coding agents).

An MCP server bridges that gap. The LLM agent decides when to call the tool and passes the state; the server sends a request to TypeSafe's single evaluation endpoint, POST https://api.typesafe.ai/v1/systemone, and hands the typed answer back (API reference). This is the smallest request the API reference shows:

{
  "state": "Help! My payouts have been failing for 3 days.",
  "model": "jev-latest",
  "questions": {
    "is_urgent": {
      "type": "noul",
      "instructions": "Does this convey urgency?"
    }
  }
}

In community servers, tools with names like classify, verify or screen wrap requests like this one with preset questions.

Why Jev works well as an MCP tool

• Typed by construction. TypeSafe suggests Jev where you would otherwise ask an LLM to "return JSON", because the call returns typed values directly (Jev with coding agents).
• Many questions, one call. Every question is evaluated in parallel and in isolation against the same state, and adding questions barely changes response time (Introduction).
• Confidence the agent can act on. Choice and Score answers carry a confidence from 0 to 1, so the caller can act, ask for confirmation or hand off to a person (Confidence).
• Cheap enough to call often. Jev 1.13 costs $0.042 per million input tokens, and output tokens are free (Models). See Jev pricing for more.

TypeSafe's launch post calls Jev "a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out" (launch post). That is exactly the shape of an MCP tool. For how this differs from asking the agent's own model, see Jev vs an LLM.

What to check before you plug one in

• Input limits. Jev takes text only, with 64k tokens per request and 32k for the state plus the longest question (Models). Large files or screenshots need converting to text first.
• Rate limits. Jev 1.13 allows 1,200 requests per minute, and a request over the limit returns 429 Too Many Requests (Models). An agent that calls a tool in a loop can hit this.
• Model version. The jev-latest alias moves when a new release ships. If a server's thresholds were tuned on one version, pin that version ID (Models).
• Where decisions live. TypeSafe advises keeping questions and threshold constants in one file so people can review them (Agent skill). A good MCP server makes those easy to find and change.

If the server gates risky tool calls, also read safety and guardrails. For browser-driving servers, see agents and browsers.

Projects doing this

SDKs & integrationsjev-mcp (jkudish)An npm MCP server that exposes ten judgment tools built on Jev, such as verify, screen, rerank, classify and gate, instead of one raw evaluate call.GitHub · ★ 320 · jkudishSDKs & integrationsTypeSafe MCPA Go CLI and single-binary stdio MCP server that gives Claude Desktop, Claude Code and Codex direct access to Jev judgments, with a pi extension too.GitHub · ★ 304 · itsmostafaSDKs & integrationsjev-studioA Python kit with a jev CLI and an MCP server that exposes Choice, Noul and Score tools plus a slash command per TypeSafe cookbook.GitHub · ★ 15 · utk2103SDKs & integrationsjev-mcp (burnigtm)An MCP server that puts Jev on the coding loop in Cursor, Codex and any other MCP client.GitHub · ★ 43 · burnigtmSDKs & integrationsJevbridgeAn ACP and MCP adapter that brings Jev decisions into harnesses such as Codex, Claude, Grok and OpenCode without replacing them.GitHub · ★ 41 · tacticoccSDKs & integrationsjev-judge-mcpPython MCP tools for verify, screen, classify, rerank and more, where Jev judges and a policy decides whether to act, review or escalate.GitHub · ★ 24 · PyModelSDKs & integrationsjev-mcp (blakestone-x)An MCP server offering typed classify, score, check, match and screen tools for any agent, with confidence on every answer.GitHub · ★ 22 · blakestone-xSDKs & integrationsjevwireA decision layer for agents that ships as an MCP server, an embeddable library and an escalate-only Claude Code plugin.GitHub · ★ 18 · BrainwiresSDKs & integrationsjev-cli (tumf)An unofficial, dependency-light Python CLI and stdio MCP server for asking Jev noul, choice and score questions and getting machine-readable answers.GitHub · ★ 13 · tumfSDKs & integrationsjev-mcp-serverA Python MCP server that exposes Jev's three question types plus compare, verify and batch classify, with a one-command client installer.GitHub · ★ 2 · wangkuangkuangSDKs & integrationsjev-mcp (freepik-company)A dependency-light Go MCP server exposing decide, classify, verify, rerank and list_models tools over Jev.GitHub · ★ 2 · freepik-company

Questions

More guides: What is Jev?Jev pricing and API costJev vs LLMs: when to use whichHow to use Jev: a quickstartHow to use Jev with Claude CodeJev in an agentic harness: where it fits in an agent loopJev for orchestration: routing requests, models and skillsJev as a judge: evaluating LLM and agent outputsJev for SEO and GEOJev for ads and ad reviewJev for marketing and lead scoringIs there an open source Jev?Can you fine-tune Jev?Designing questions for JevWhat is RLCD?Jev statistics