openjev
opencode plugin that routes the agent's small multiple-answer decisions to Jev and labels each answer auto or escalate by confidence.
Overview
OpenJev for opencode (package opencode-openjev) gives the agent a decision tool for questions with few possible answers, such as whether a command is safe, which tool fits, whether a PR is ready to merge or how serious a bug is. You pass the text and possible answers; Jev returns which one fits and how confident it is, and the plugin labels every answer auto (confident enough to act) or escalate (send to the LLM or a person).
The README reports measured p50 latency of 267 ms, about $0.017 per 1,000 decisions, up to 32 questions per call with 27 questions taking about 5 ms longer than one, and that choice answers scored 0.75 or higher were right 96% of the time in its tests. A support-ticket example routes a clear billing issue automatically at 0.99 and flags a vague one at 0.38 for a person. It is MIT licensed with a five-minute setup guide.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this command safe to run? | yes · no |
| Choice | Which tool fits this task? | — |
| Score | How serious is this bug? | — |
| Choice | Which team should handle this ticket? | billing · technical · sales · spam |
How it works
- Agent calls the OpenJev tool with text and options.
- Jev returns the chosen option and confidence.
- Plugin labels the answer auto or escalate.
- Agent acts or hands off to the LLM or a person.
Strengths
- Sub-second decisions at low cost per the README.
- Calibrated confidence drives auto/escalate.
- Many questions per call.
Caveats
- Requires a TypeSafe key.
- Numbers come from the project's own tests.
Getting started
Follow SETUP.md to install opencode-openjev (Node 20+, opencode 1.18+) and add your TypeSafe key.At a glance
- Type
- Plugin
- Status
- beta
- Language
- TypeScript
- Runtime
- opencode plugin, Node
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
OpenJev: An Opencode plugin that replaces text-generation decisions with Jev (TypeSafe System One).
logicrw/awesome-jev-projectsOpenJev: An Opencode plugin that replaces text-generation decisions with Jev…
wh000wh000/awesome-jev-live
Frequently asked questions
What does auto vs escalate mean?
Auto answers are confident enough to act on; escalate answers go to the LLM or a person.
How much does it cost?
About $0.017 per 1,000 decisions per the README.
How accurate are confident answers?
Choice answers at 0.75 or higher were right 96% of the time in the author's tests.
Related projects
DeepSeek Harness plugin with 12 opt-in Jev or Luna Decisions judgments: skill and file selection, supervision, filtering and approval help.
PluginbetaTypeScriptDeepSeek Harness jev_decide tool that brings the Jev decision paradigm to any OpenAI-compatible LLM, or to real Jev with a key.
ProjectprototypeTypeScriptSix-file DSPy coding agent where Jev answers closed questions (safety, tool fit, history, delegation) and Python owns policy.
ProjectprototypePythonSkill and CLI that compares a shortlist of agents, MCP servers or skills against a task and permission ceiling, abstaining on weak leads.
MCP serverprototypeTypeScriptStrict-allowlist shell harness where Jev walks a TOML decision tree to pick a bash command, flags and paths, confirming writes.
ProjectprototypePythonRust template for multi-agent orchestration with Jev as the L0 layer for typed routing, risk and escalation decisions.
ProjectprototypeRust