Jev DSH Decision Engine
Decision plugin for agent harnesses where Jev recommends tools, skills or owners and scores output, for DeepSeek Harness and iPolloWork.
Overview
Jev DSH Decision Engine brings Jev's structured decisions into agent harnesses. It helps choose among the tools, skills and agents that are actually available and permitted, scores output against explicit criteria, judges whether a capability is needed, and batches independent questions. It returns choices, probabilities, confidence and usage, but never executes the recommendation or reroutes all requests; the original agent keeps planning, explanation and execution.
It ships as a native DeepSeek Harness plugin with jev_status, jev_check_connection and jev_evaluate tools plus a jev-decision skill, and as an iPolloWork import package that serves OpenCode, DeepSeek Harness and Codex Harness through one install. Documentation is in Chinese. Users supply their own TYPESAFE_API_KEY through the host's credential store, and connection tests incur small API usage.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which available tool, skill or agent should handle this task? | — |
| Score | How well does this output meet the stated criteria? | — |
| Yes / no | Is this capability needed for the task? | — |
How it works
- Install the plugin in DeepSeek Harness or iPolloWork
- Configure TYPESAFE_API_KEY in host credentials
- Agent calls jev_evaluate with candidates and criteria
- Agent uses the returned choice and probabilities; Jev executes nothing
Strengths
- One decision core for several harnesses
- Recommends only from actually available, permitted tools
- Does not take over the main agent
Caveats
- OpenCode and Codex support only via iPolloWork
- Chinese documentation
Getting started
Clone the repo and run `dsh plugin add /absolute/path/jev-dsh-decision`, then set TYPESAFE_API_KEY in DSH credentials.At a glance
- Type
- Agent skill
- Status
- beta
- Language
- TypeScript
- Runtime
- DeepSeek Harness plugin and iPolloWork extension
- Has code
- yes
Use cases and integrations
How curated lists describe it
Provides Jev structured decision support for Agent Harness to recommend tools, Skills and Agents and return judgments with probabilities, with a native DeepSeek Harness plugin and an iPolloWork entry serving OpenCode, DeepSeek Harness and Codex Harness.
logicrw/awesome-jev-projectsJev DSH 决策引擎｜面向 Agent Harness 的结构化决策插件。原生支持 DeepSeek Harness，通过 iPolloWork 支持 OpenCode、Codex Harness。
hellogumbo/awesome-jevDecision plugin for agent harnesses: Jev picks the tool, skill or owner and scores the output, while the agent keeps planning and execution. Ships for DeepSeek Harness and iPolloWork.
MrJev/awesome-jev
Frequently asked questions
What does Jev decide?
Which tool, skill or agent fits a task, and how output scores against criteria.
Does it execute actions?
No, it only returns recommendations with probabilities.
Which harnesses are supported?
DeepSeek Harness natively, plus OpenCode and Codex Harness via iPolloWork.
Related projects
Local MCP server putting Jev in the coding loop: route next steps, pick prepared tool calls, review patches and verify claims.
MCP serverbetaTypeScriptDeepSeek Harness plugin that registers a native jev_decide tool and declares its own policy for when the agent should ask Jev.
PluginbetaJavaScriptCodex plugin that consults Jev to choose tools, models and skills, diagnose failed commands and check completion claims.
PluginbetaTypeScriptMCP server exposing Jev Noul, Choice and Score via OpenRouter so coding agents can make quick decisions without reasoning tokens.
MCP serverbetaTypeScriptPortable decision layer for agent harnesses: Jev picks from host-supplied capabilities while the host keeps permissions and execution.
ProjectbetaTypeScriptEval-first MCP server for Jev: prototype questions, map them over many items, and measure accuracy and thresholds on labelled data.
MCP serverbetaTypeScript