Article
CommunityJev Guides & Articles
Jev Usage Walkthrough: API, SDK, and jev-review
A step-by-step tutorial by 雪庭 that goes from a first curl call against api.typesafe.ai through the Python typesafe-sdk to the jev-review MCP plugin for Claude Code and Codex, including typed Choice, Score, and Noul questions in one request.
- Format
- Article
- Reading time
- 2 min
- Published by
- Community
- Author
- 雪庭
- Last verified
- Oct 2026
Highlights
- The SDK example sends intent (Choice), urgency (Score), and has_deadline (Noul) in a single client.system_one call.
- A curl call to api.typesafe.ai/v1/systemone returns model jev-1.13.0 with answers.has_deadline.noul at 0.83 for a refund complaint.
- jev-review installs via npx plugins add NiazMorshed2007/jev-review and reads JEV_API_KEY, not TYPESAFE_API_KEY.
- A sample review of a documentation commit scored correctness 4.7 at 0.55 confidence and readability 5.4 at 0.37 confidence.
- The jev-model-router plugin needs Claude Code 2.1.259 or newer with CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1.
Watch out
No licence is stated for jev-review or jev-model-router, and the published review scores come from a single documentation commit. The walkthrough needs a valid TypeSafe key and quotes no pricing, and the router plugin sits behind an early-access function-hooks flag whose API may change in later versions.
Questions about this entry
Short answers drawn from the linked project or page.
A working Jev judgment pipeline: a first curl call against api.typesafe.ai, a Python typesafe-sdk script mixing Choice, Score, and Noul questions, and the jev-review MCP plugin running inside Claude Code or Codex.
Claude Code or Codex already working, a TypeSafe API key, Node.js 20 or newer for the community review plugin, Python 3.10 or newer for the SDK examples, and a macOS, Linux, or WSL terminal with the example commands followed as written.
The sample asks intent as a Choice with refund, technical, information, and other criteria; urgency as a three-level Score from no deadline to immediate impact; and has_deadline as a Noul, all inside one client.system_one call.
It is a statistic computed from the answer's probability distribution, and the article warns it is not the probability that the answer is correct. Noul answers have no separate confidence field, so a returned probability must be read as a calibrated truth value.
Run npx plugins add NiazMorshed2007/jev-review with --target claude-code or --target codex, set JEV_API_KEY separately from TYPESAFE_API_KEY, restart the client, and confirm the MCP connection, for example with /mcp in Claude Code.
For a documentation-only commit, Jev returned scalar scores with confidences: correctness 4.7 at 0.55, consistency 4.9 at 0.55, documentation 4.9 at 0.53, maintainability 5.0 at 0.54, and readability 5.4 at 0.37.
The article says the common mistake is compressing everything into one vague question such as whether something is reasonable. Define each condition separately, keep an other option in every Choice, and record the model identifier from each response.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: jev-review: a local-first score loop for coding agents
X: Jev in front of the HeyGen MCP
Hacker News: Thread on Selmar's Jev code-review calibration notes and ordering.py comparisons.
Lectem2 pointsCalibrating Jev as a Code ReviewerRead the thread on Hacker NewsX: Playwright CLI + Jev vs Playwright MCP
More Jev resources like this
accept99 explains how Jev differs from chat models and how to run Jev-like decision models for free: Ollama serves nimble 9B and tev1 4B and 0.8B over /v1/systemone, while the open-source Laya project offers pip-installable weights and an MLX build for Macs.
Jev Guides & ArticlesArticle#community#article#local
Community
Code Mode for MCP where the sub-model is a calibrated decision model instead of an LLM: agents get search(query) and execute(code), and Jev makes per-item choices in about 100 ms, benchmarked on MCPToolBench++, LiveMCPBench and When2Call.
Jev Agents & MCP#community#catalog-import#logicrw
Community
vishiatchoudhary/tooljevPython
vishiatchoudhary/tooljevPythonEvaluation scripts, MCP tools, and a skill router built to measure where jev-1.13.0 fits in agent harnesses, with raw results under eval/.
Jev Benchmarks & Evaluations#community#evaluation#agents
Community
Aitejiu/jev-harness-labPython
Aitejiu/jev-harness-labPythonwangkuangkuang's Python MCP bridge to Jev: three official question types plus compare, verify, and batch classify, with a one-command installer that writes client config for Claude Code, Pi, Cursor, OpenCode, or Codex.
Jev Agents & MCP#community#mcp#python
Community
MCP.so's explainer on Jev: how Choice, Score, and Noul differ from text generation, where a decision layer fits in retrieval, verification, routing, and ranking, and the four tools in the community Jev MCP server.
Jev Guides & ArticlesArticle#community#mcp#article
Community
Latitude is an open-source observability platform for AI agents: one line of telemetry captures traces, failing traces group into tracked signals, and fixes are dispatched to coding agents and replayed against real traces.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
latitude-dev/latitude-llmTypeScript
latitude-dev/latitude-llmTypeScriptFrom 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.