Jev directory
CommunityJev Agents & MCP
Jev Decision MCP
Amidabuddha's local stdio MCP server wraps TypeSafe's Jev in a single jev_decide tool: batched choice, score, and noul judgments with probabilities and token usage, leaving thresholds and actions to the caller.
- Reading time
- 2 min
- Published by
- Community
- Author
- amidabuddha
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- TypeScript
- Capabilities
- Open sourceRequires API key
Highlights
- One read-only tool, jev_decide, over stdio; it returns judgments and usage and never executes actions.
- choice takes 1-255 named options; score rates 2-10 ordered levels with a 0-based score; noul returns a 0-1 probability.
- Batches independent questions in one call and preserves question IDs; questions in a batch cannot see each other's answers.
- Runs via npx -y [email protected], a global npm install, or a source build with npm ci and npm run build.
- Env: TYPESAFE_API_KEY, TYPESAFE_DEFAULT_MODEL (jev-latest default), and JEV_TIMEOUT_MS (30000 ms default, up to 120000).
Watch out
Independent community project, not affiliated with TypeSafe; it needs a TypeSafe key for actual calls, and without one the tool still appears but calls return MISSING_API_KEY. The npm install does not read a .env from the caller's working directory, and upstream error bodies are not exposed.
Questions about this entry
Short answers drawn from the linked project or page.
Add an mcpServers entry that runs npx -y [email protected] with TYPESAFE_API_KEY in its environment, then restart or reload your client's MCP servers. The tool appears as jev_decide; if the host cannot resolve npx, use its platform-specific launcher or an absolute path.
Model, answers, and token usage as both MCP structured content and JSON text. It accepts a state containing relevant text or JSON plus named independent questions, batching them in one call while preserving question IDs, and it never executes actions.
choice picks a label from 1-255 options with probabilities and confidence; score rates one dimension on 2-10 ordered levels, returning a 0-based score, legend, probabilities, and confidence; noul returns the probability that a condition holds, from 0 to 1.
No. It leaves thresholds, escalation, and action execution to the caller; it only returns judgments and usage. Confidence does not authorize actions or guarantee correctness, and thresholds should be evaluated on representative data.
Set TYPESAFE_API_KEY in the MCP client's server environment, or in .env for a source build (the npm install ignores a caller-directory .env). The server starts without a key so a client can discover the tool, but decision calls then return MISSING_API_KEY.
Yes. Each call can override the model, and TYPESAFE_DEFAULT_MODEL sets the default (jev-latest). The listing warns that jev-latest follows TypeSafe model updates, so set a specific supported version for repeatable evaluations, and keep the host tool timeout above JEV_TIMEOUT_MS divided by 1000.
No. It is an independent community project written in TypeScript with the official MCP and TypeSafe SDKs, licensed under MIT, and it exposes the official TypeSafe API rather than reimplementing Jev.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev in front of the HeyGen MCP
X: Playwright CLI + Jev vs Playwright MCP
X: A local MCP server that cuts computer use from 23s to 0.8s. Cartwright built Cortex, a local MCP server where…
X: A side panel that drives any site's WebMCP tools
More Jev resources like this
Five typed tools — classify, check, score, rank, and ask — plus a skill, wired into Claude Code, Codex, Pi, and OpenCode in one setup command.
Jev Agents & MCP#community#typescript#mcp
Community
FrancoisChastel/jev-codeTypeScript
FrancoisChastel/jev-codeTypeScriptDevMortimer's Pi extension hands batched Choice, Score, and Noul judgments to Jev: a typesafe_evaluate agent tool, a terminal playground, and a typed API other extensions reuse, with daily request, token, and dollar caps.
Jev Agents & MCP#community#pi#typescript
Community
Hands the agent's browser clicks and typing to Jev: wingman hides Playwright MCP's page-action tools, asks Jev which element fits each step, and returns the step to the agent when it is unsure.
Jev Agents & MCP#community#typescript#mcp
A single-binary MCP tool named evaluate that sends Noul, Choice, and Score questions to hosted Jev or a local System One server.
Jev Agents & MCP#community#go#mcp
Community
itsmostafa/system-one-connectorGo
itsmostafa/system-one-connectorGoA very WIP tool for jev-ification from Harry Solovay: lessons from coding sessions become approved Markdown traits, and a configurable decision model evaluates code one trait at a time while an MCP server handles group-aware approval.
Jev Agents & MCP#community#github#typescript
Community
harrysolovay/traits-pocTypeScript
harrysolovay/traits-pocTypeScriptAn independent agent skill that walks a coding agent through designing a bounded Jev decision layer: contract, criteria, error behavior, fallback, and evaluation plan first, then scaffolding through TypeSafe or OpenRouter.
Jev Agents & MCP#community#catalog-import#logicrw
Community
grandamenium/jev-anythingPython
grandamenium/jev-anythingPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.