User repository on GitHub
Jev周りの環境構築や検証，ツール化用
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
Jev tooling (ishigaki0302)
A Japanese working repository that gathers Jev environment setup, notes and tooling: an MCP server pinned to clef:flash, a Claude Code skill for deciding when to delegate to a decision model, call logs and a small evaluation set of delegation tasks.
- Stars
- 0
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- ishigaki0302
- Last verified
- Oct 2026
Highlights
- MCP server mcp/jev_mcp.py (standard library only) exposes mcp__jev__decide for typed choice / score / noul questions on one or many states, plus a status tool.
- Model pinned to clef:flash and kept in memory for 30 minutes via JEV_MODEL and JEV_KEEP_ALIVE; 3-6 seconds per decision on CPU.
- A Claude Code skill in .claude/skills/jev/ tells Claude when to delegate work to Jev, linked from the user's global skills folder.
- Every call is logged one JSONL line at a time to logs/calls.jsonl; show_log.py aggregates, filters low-confidence verdicts or prints the last call in full.
- The README also documents the installed Ollaya version, notes and an eval/ set of delegation-candidate tasks.
Quickstart
bash
python3 mcp/show_log.py
python3 mcp/show_log.py --low
python3 mcp/show_log.py --full 1Watch out
README reviewed 2026-10-17 from the round-eleven capture, which is truncated at the uninstall steps: exact model versions, eval results, licence terms (none declared on GitHub) and the roughly 20 s cold-load and 19 GB memory figures are the author's own.
Questions about this entry
Short answers drawn from the linked project or page.
A Japanese working repository for Jev: environment setup, investigation notes, an MCP server, a Claude Code skill, call logs and an evaluation set, organized as JEV.md, NOTES.md, LOG.md and code under mcp/ and eval/.
mcp__jev__decide takes a single state or multiple states and asks typed questions — choice, score or noul — for a verdict; mcp__jev__status shows the local server's status and the loaded model.
The server is pinned to clef:flash and keeps it in memory for 30 minutes, adjustable with JEV_MODEL and JEV_KEEP_ALIVE variables. The README documents the installed runtime version in LOG.md.
Three to six seconds per item on CPU, with roughly 20 seconds of loading after idle and about 19 GB of memory used, all per the author's own README measurements.
No. The README says it is decision-only: it delegates typed decisions and cannot summarize or generate prose, which is why the skill is about choosing when to delegate.
Run python3 mcp/show_log.py for an aggregate and the last 20 calls, add --low for low-confidence verdicts only, or --full 1 to print the most recent call in full. Logs stay in logs/calls.jsonl, kept out of git.
It treats Jev as the decision-model concept and pins a local clef:flash model; the captured digest shows no TypeSafe endpoint or key. Details are in the Japanese README and JEV.md.
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
Djaevin, a calibrated, local-first decision layer from Icaro0310 that runs behind a Devin ACP backend with no API key. It ships as an MCP server published in the official MCP Registry and installs from PyPI as poordjaevin; the repo topics name calibration, conformal prediction and NLI.
Jev Agents & MCP#community#github#python
Community
icaro0310/devin-judgePython
icaro0310/devin-judgePythonA plugin-based agent harness (FastAPI + Vue 3) bundling a chat GUI, multi-provider model access, an MCP client and marketplace, skills, workflow studio, sandboxed execution and a Tauri desktop shell. It integrates TypeSafe AI's Jev for structured Choice, Score and Noul decisions.
Jev Agents & MCP#community#github#agents
Community
andyqiuqiubo/andy-harnessTypeScript
andyqiuqiubo/andy-harnessTypeScriptAn open-source computer-use MCP server for Claude Code, Codex and other agents that reads the accessibility tree instead of screenshotting every step, using Jev as a fast judge and, experimentally, as image input to spot icons and game objects.
Jev Agents & MCP#community#github#mcp
Community
8itlab/reflex-cuPython
8itlab/reflex-cuPythonA lightweight self-hosted System 1 decision daemon and MCP server for AI agents and multi-agent swarms: it evaluates state text against typed choice, score and noul questions in one forward pass, with Laya, Ollama and OpenAI-compatible backends.
Jev Agents & MCP#community#github#mcp
Community
axiomantic/local-systemonePython
axiomantic/local-systemonePythonAn open-source Jev-based MCP router: it discovers, catalogs, classifies and deduplicates every MCP server and tool you have, then exposes each agent only the small, authorized subset it needs for the task. Routing decisions are System One Choice/Score/Noul questions, scored by a pluggable engine at the edge or hosted.
Jev Agents & MCP#community#catalog-import#logicrw
Community
amantux/mcp-routerPython
amantux/mcp-routerPythonReusable 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/aiwgTypeScriptFrom 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.