User repository on GitHub
Just Jev it. A Jev-first agent loop with reusable actions for DeepSeek Harness.
- Forks
- 2
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
JevDo
A Jev-first agent loop for DeepSeek Harness: Jev reads the same assembled session and tool history as the main model and either dispatches a saved Action, asks the main model, clarifies, or ends the turn. Requests covered by verified Actions can finish with zero main-model calls.
- Stars
- 27
- Language
- TypeScript
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- yinhong-zhou
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- TypeScript
- Capabilities
- Open source
Highlights
- Jev reads the same assembled session and tool history as the main model and picks the next step: run an Action, ask the main model, clarify, or end.
- Actions are commands or scripts with purpose, project binding and acceptance checks, kept across sessions and validated before reuse.
- README reports a full start to code change to test/build handoff ran in the project's experiments; covered requests can run with zero main-model calls.
- Package and tool prefix keep the historical name jevaction; early builds target DeepSeek Harness 0.2.0-rc.1 for headless use.
- MIT TypeScript repo on Node.js 24.14+: 27 stars, 2 forks, created 2026-10-02 and last pushed 2026-10-03, so it is an early-stage project.
Watch out
README reviewed 2026-10-17; it is mostly Chinese and the digest truncates inside its workflow diagram, so install commands, Action validation details and configuration are not covered here. The zero-main-model-call claim comes from the project's own experiment reports, not independent testing.
Questions about this entry
Short answers drawn from the linked project or page.
JevDo is a Jev-first agent loop for DeepSeek Harness: the model is treated as one tool among several, and Jev decides the next step from the same session and tool history the main model sees.
Jev reads the assembled conversation and tool history and either schedules an Action, hands the request to the main model, asks for clarification, or ends the turn; the main model is an option, not the driver.
A reusable command or script plus its purpose, project binding and acceptance checks, persisted across sessions. Simple tasks are saved as commands, and complex behaviour is better stored as a script.
Yes, when an existing Action covers all the work and its completion criteria are met — the README gives a plain project start or a full regression as examples — the README says Jev can complete with zero main-model calls.
Node.js 24.14 or newer and DeepSeek Harness, with early builds targeting DSH 0.2.0-rc.1 headless use; a main model handles new problems and exceptions. The package and tool prefix keep the name jevaction.
A normal loop asks the model what to do next every time. JevDo asks Jev first, so repeated operations are dispatched straight to validated Actions and the main model is reserved for new problems.
yinhong-zhou published it under MIT on 2026-10-02; the digest shows 27 stars, 2 forks and a push on 2026-10-03, so expect an early project whose evidence lives in the repository's own experiment reports.
More Jev resources like this
A developer-preview personal agent harness that turns a sequence of model calls into continuing work: understand the request, choose an action, execute, inspect evidence and recover context. You pick the general model and can optionally connect a Reflex/Jev decision service.
Jev Agents & MCP#community#github#agent
A Claude Code plugin that sends every Agent spawn through JEV (TypeSafe), which picks the model (haiku, sonnet, opus), the effort and the tool profile of each subagent. It fails open to sonnet at medium effort, accepts a TypeSafe or OpenRouter key, and ships plan and run skills.
Jev Agents & MCP#community#github#claude-code
Community
hiyoku/jev-orcTypeScript
hiyoku/jev-orcTypeScriptA coding agent for your terminal: you chat, and when you ask for a change it reads, edits and runs your code itself. Jev helps on the side with a few quick calls that spot a simple question so replies come back faster, and its key is optional.
Jev Agents & MCP#community#catalog-import#logicrw
Community
coasty-ai/jevcodeTypeScript
coasty-ai/jevcodeTypeScriptA Claude Code plugin that hands Claude Code's small, frequent decisions to Jev at six points: four event hooks and two routing decisions, each a roughly one-second call that reduces permission prompts, warns when fetched content addresses the agent, re-checks done, and routes effort and model.
Jev Agents & MCP#community#github#claude-code
Community
yinjs/claude-jevTypeScript
yinjs/claude-jevTypeScriptA Claude Code configuration CLI shipping Jev modules for model routing, skill suggestion, tool guardrails, auto-mode gates, and sandbox hooks.
Jev Agents & MCP#community#typescript#coding-agent
Community
davila7/claude-code-templatesPython
davila7/claude-code-templatesPythonAn experimental coding agent CLI built on opencode whose v2.0.0 architecture adds a JEV decision layer with five decision gates, shadow mode, and ECE calibration, plus a three-level RSI self-evolution loop and six Pi-framework extensions.
Jev Agents & MCP#community#github#typescript
Community
momozi1996/momo-codeTypeScript
momozi1996/momo-codeTypeScriptFrom 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.