Jev directory
CommunityJev Agents & MCP
Decisionkit — npmjs.com
Provider-neutral CLI, Claude Code plugin, and MCP server that asks yes/no, choice, score, or rank questions about text and returns probabilities. Talks to Jev, Jev via OpenRouter, the OpenAI Decisions API, or Claude.
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
This is one of several Jev listings for the same project. The canonical entry is Decisionkit — adityajha2005.
Highlights
- Supports binary, choice (2–255 options), score (2–10 levels), and rank (2–20 candidates) questions.
- Talks to Jev, Jev via OpenRouter, the OpenAI Decisions API, or Claude through an existing Claude Code login.
- In a 100-task benchmark, Jev via OpenRouter scored 97/100 at $0.0018 per 100 decisions with 524 ms median latency.
- Requires Node.js 20+; works as a CLI, Claude Code plugin, or MCP server for Codex, Cursor, VS Code, Gemini CLI, Claude Desktop, Windsurf, and Zed.
- Errors, refusals, and invalid provider responses are always returned as errors, never as answers.
Questions about this entry
Short answers drawn from the linked project or page.
Run npm i -g @adxtya21/decisionkit. The npm package is @adxtya21/decisionkit and the command it installs is decisionkit. Node.js 20 or later must be on your PATH.
Four: Claude through the Claude Code login you already have, the OpenAI Decisions API at POST api.openai.com/v1/decisions, Jev by TypeSafe AI at POST api.typesafe.ai/v1/systemone, and Jev via OpenRouter at POST openrouter.ai/api/v1/systemone.
binary returns probability_true; choice takes 2–255 options and returns a probability per option; score takes 2–10 levels and returns expected_level, most_likely, and a probability per level; rank takes 2–20 candidates and sorts them by score.
Each decision sends the context and questions only to the one provider you configured, and nothing is sent anywhere else or automatically. Keep secrets, credentials, personal data, and uncleared code out of context. No prompt content is logged.
Not if you set DECISIONKIT_PROVIDER=claude, which runs on your Claude Code login and strips ANTHROPIC_API_KEY from its own process. Otherwise set DECISIONKIT_PROVIDER plus the matching DECISIONKIT_* key. Only DECISIONKIT_* variables are read.
No. They are the provider's own estimates and have not been verified as calibrated for your task, so you should pick thresholds using your own labeled examples. Neither provider returns a rationale, and decisionkit never invents one.
Rank is derived from one score per candidate rather than a direct comparison, so close values should be treated as ties. Provider-specific fields such as confidence are kept under native rather than normalized.
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
Standalone MCP server that exposes Jev's typed decisions to MCP hosts such as Cursor and Claude Code, with an HTTP mode and an importable library. Handles choice, score, and boolean questions across Vercel, OpenRouter, and TypeSafe providers.
Jev Agents & MCP#mcp#typescript#openrouter
Community
An MCP server that exposes TypeSafe's Jev model as twelve judgment tools, including claim verification, injection screening, reranking, classification, extraction, and patch review, each returning typed probabilities in roughly 150 to 500 ms.
Jev Agents & MCP#mcp#agents#typescript
Community
A Model Context Protocol server that exposes TypeSafe AI's Jev decision model as typed tools over stdio, runnable with npx or bunx. Ships jev_decide, jev_choice, jev_score, jev_noul, jev_box and jev_list_models for MCP clients such as Claude Code, Cursor and VS Code.
Jev Agents & MCP#mcp#typescript#agents
Communitydevbux/typesafe-mcp
Reusable project context and specialist workflows for the AI tools you already use: agents, skills, rules, and MCP tools for Claude Code, Codex, Copilot, and Cursor, plus Jev decision support that offloads bounded LLM decisions to save tokens.
Jev Agents & MCP#community#github#agents
Community
jmagly/aiwgTypeScript
jmagly/aiwgTypeScriptA Claude Code plugin and npm library that replaces the LLM compaction summary with Jev decisions: each tool call and result is scored, stale ones are dropped or truncated, and everything kept stays verbatim. A fork adds jev-compact and jev-gate CLIs that run through OpenRouter.
Jev Agents & MCP#claude-code#coding-agent#cli
Community
A community Go CLI and MCP server that puts Jev's native decision API into coding-agent workflows. It runs noul, choice, score, box, and systemone decisions from the shell or as MCP tools, with dry-run request previews, caller-supplied acceptance gates, and ordered JSONL batches.
Jev Agents & MCP#mcp#cli#agents
Communitykoltyakov/decide
From 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.