Skip to content

Jev Workbench

Local service for authoring versioned judgment functions on TypeSafe Jev in a browser, then publishing one immutable version that both your backend calls over HTTP and your coding agents call over MCP, using a scoped client token.

Stars
2
Language
TypeScript
License
MIT
Published by
Community
Author
molis-ai
Last verified
Oct 2026

molis-ai/jev-workbench

Organization repository on GitHub

View on GitHub

Build versioned judgment functions on TypeSafe's Jev once, then call the same published version from your backend over HTTP and from coding agents over MCP. The vendor key stays on your machine.

Forks
1
Last pushed
Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Write a judgment once. Call it from everywhere. Jev Workbench is a local service for building judgment functions on top of TypeSafe's Jev — small, versioned decisions like "is this ticket about billing?" or "does this evidence support that claim?". You define one in the browser, try it on real input, and publish it. From then on the same published version answers both your backend over HTTP and your coding agents over MCP.

Quoted from the project's README. Read the full README.

Highlights

  • Three primitives: Noul returns a probability of yes, Choice returns a key plus a probability per option, Score returns a probability-weighted level index.
  • Publishing freezes the config and model version and requires a successful preview of exactly that config; published versions never change under a caller.
  • Grants pin a client token to specific function versions, so publishing v2 does not silently upgrade anyone.
  • Run records keep status, timing, model and error code only; input text and answers are not stored, saved test cases excepted.

Watch out

Requires Node.js 24 and pnpm 11.9.0. The demo on port 17430 serves simulated answers with no network and is not evidence that a production publish will work. Without a TypeSafe key you can still edit everything but inference returns PROVIDER_NOT_CONFIGURED, and production never falls back to simulated answers. It binds 127.0.0.1:17420 only. POST /v1/systemone and GET /v1/models need an explicit Allow official Jev calls setting, and the official Python SDK works against it only if the token has that access. Each agent install writes its own credential file at mode 0600, and uninstall stops if the entry was hand-edited. 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.

More Jev resources like this

22GitHub stars
Runnable agent harness that moves pick, score and yes-no loop choices to Jev or Laya in milliseconds while reserving the language model for writing, compiling DECISION.md into executable questions.
Jev Agents & MCP#community#jevbest#catalog-import
15GitHub stars
Node CLI that lints a file tree against fuzzy natural-language rules in one Markdown file, judging each matched file with a swappable TypeSafe-backed judge.
Jev Agents & MCP#community#jevbest#catalog-import
22GitHub stars
Command-line toolkit that wraps Jev judgments into scriptable verify, screen, classify, extract, find, rerank, match, route, ask, compact, and batch commands.
Jev Agents & MCP#community#jevbest#catalog-import
24GitHub stars
Wires Jev into agents as seven MCP tools, an embeddable JevDecisionModel library with run functions, and Claude Code hooks at tool and turn boundaries.
Jev Agents & MCP#community#jevbest#catalog-import
12GitHub stars
Pi extension that puts Jev behind coding-agent hooks: JSON packs ask closed questions on redacted state and turn probabilities into block, confirm, annotate, or steer actions.
Jev Agents & MCP#community#jevbest#catalog-import
7GitHub stars
MCP server exposing Jev Choice, Score, Noul, batched ask, file-reading triage, and model listing as typed tools with full probability distributions.
Jev Agents & MCP#community#jevbest#catalog-import
Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai