User repository on GitHub
MCP server exposing TypeSafe's Jev model as a typed evaluate tool.
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Practices & Patterns
TypeSafe Jev MCP
Single-tool MCP server for Jev: send state plus a map of noul, choice or score questions to evaluate and get typed answers with probabilities and confidence back.
- Stars
- 3
- Language
- Rust
- License
- MIT
- Published by
- Community
- Author
- anasbekheit
- Last verified
- Oct 2026
About this project
MCP server for TypeSafe Jev. One tool, evaluate: state + typed questions in, noul / choice / score with probabilities out. Not affiliated with TypeSafe. Needs Claude Code, Codex, OpenCode, Antigravity, or Cursor. Opens console.typesafe.ai or prints the URL; paste a key. No TTY: TYPESAFEAPIKEY='…' and re-run. Restart the agent. Binary only: releases, npx typesafe-jev-mcp, cargo install typesafe-jev-mcp. Live docs: docs.typesafe.ai/llms.txt. Pin a versioned model (jev-1.13.0) after you tune a threshold.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
- Languages
- Rust
Highlights
- One tool, evaluate: state plus typed questions in, noul, choice and score answers with probabilities out.
- Installs with a curl-piped shell script on macOS and Linux or a PowerShell one-liner on Windows.
- Binaries also ship through GitHub releases, npx typesafe-jev-mcp and cargo install, if you would rather not pipe a script.
- Setup opens console.typesafe.ai or prints the URL for a key, and falls back to TYPESAFE_API_KEY when there is no TTY.
Watch out
The documented install path pipes a remote script straight into sh or iex, which is worth reading first if that is not your habit; the release binaries, npx and cargo install are the alternatives. The README does not name a default model, only advises pinning a versioned one such as jev-1.13.0 after you tune a threshold, and it does not document retries, caching or rate limiting. It is a thin wrapper: question design, thresholds and cost control stay in your code. It states plainly that it is not affiliated with TypeSafe, and it is MIT licensed. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
You get the three question types behind one tool plus key onboarding, and it declares support for Claude Code, Codex, OpenCode, Antigravity and Cursor. There is no batching, retry or caching layer and no decision logic. If you already have an MCP server, adding one evaluate tool is a small job and you keep control of the key.
That is your call, but you are not forced into it. The script is plain text in the repository and can be read before running. Binary-only paths exist as well: the GitHub releases, npx typesafe-jev-mcp and cargo install typesafe-jev-mcp all avoid executing fetched shell code.
The README's example returns typed answers: a noul question comes back as a number, and a choice question comes back with the chosen option, a confidence and a per-option probability map. It points at docs.typesafe.ai/llms.txt for the live contract. The example does not show a score response, so check the current docs before depending on that shape.
Tune against a versioned model and pin it. The README's advice is explicit: pin a versioned model such as jev-1.13.0 once you have tuned a threshold, and its own sample response shows jev-1.13.0. Since there is no caching or retry layer, a moved model also means every call re-bills at whatever the new version costs.
Run the install.sh (macOS/Linux) or install.ps1 (Windows) script, which opens console.typesafe.ai or prints the URL for a key, then restart the agent. Without a TTY, set TYPESAFE_API_KEY and re-run; it supports Claude Code, Codex, OpenCode, Antigravity, and Cursor.
MIT. Beyond the script you can use the release binaries, npx typesafe-jev-mcp, or cargo install typesafe-jev-mcp. It is not affiliated with TypeSafe.
One tool, evaluate, takes a state plus typed questions and returns noul, choice, or score answers with probabilities, plus the resolved model name such as jev-1.13.0.
More Jev resources like this
7GitHub stars
Python client and decision toolkit for the Jev System One endpoint with noul, choice, and score primitives, batched asks, composition helpers, concurrent evaluation, and an MCP server.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
docxology/daf-jevPython
docxology/daf-jevPython3GitHub stars
Mock carrier support agent with the guardrail layer built twice behind one interface: chat-model JSON judging versus TypeSafe Jev typed questions over the same 25 rules.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
deepansh-saxena/jev-guardrailsPython
deepansh-saxena/jev-guardrailsPython7GitHub stars
Four read-only MCP tools that rank files, cut a file down to the relevant part and pick from long lists, measured on a frozen SWE-bench Lite split.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
Bentlybro/siftrPython
Bentlybro/siftrPython3GitHub stars
Rust event bus where routing is a judgment call: one event plus one typed question per subscription goes to a Jev judge, and the calibrated probabilities decide drop, review or deliver.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
zkjoie/jevbusRust
zkjoie/jevbusRust2GitHub stars
An MCP server with two tools for agent memory: record writes facts into a per-project ledger as they happen, and recall scores each one against the task you are starting and hands back only what matters.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
Dharundp6/jev-carryforwardTypeScript
Dharundp6/jev-carryforwardTypeScript1GitHub stars
jev-filter is a CLI that filters tool output before it reaches your agent and, since 0.3, drives web pages and desktop apps too. Jev picks the next operation and its target among controls the program enumerates itself; the program re-checks and executes, and irreversible steps stop for confirmation.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
apixly-ai/jev-filterPython
apixly-ai/jev-filterPythonFrom the Jev guides
Original write-ups on what this entry is for.
Most Jev accuracy problems are state problems. What belongs in a state, what is a pre-made judgment to leave out, and why naming the slice you ask about matters more than the wording.
Jev Practices & Patterns#state#design#primitives
Read article
A Noul returns one number and no separate confidence. How to read the value, pin its boundary with criteria, split compound questions, and set a per-action threshold that matches the cost of being wrong.
Jev Practices & Patterns#noul#primitives#confidence
Read article
A Score is a weighted position, not a ruler. Why the number can land between levels, why levels are judged separately, and how to combine several atomic scores in code instead of one crowded rubric.
Jev Practices & Patterns#score#primitives#scoring
Read article
From the Jev community
Posts from builders shipping with Jev right now.