Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

athena-jev

Cognitive control layer for OpenCode that uses Jev to compute GO, DELIBERATE, VERIFY or BLOCK for agent actions, observing by default.

ProjectbetaMIT
View on GitHub ↗

Overview

ATHENA is an independent cognitive control layer for coding agents, powered by TypeSafe Jev, currently integrated with OpenCode. Its flow is: the LLM thinks, Jev judges, ATHENA observes, tools execute.

Four modules, AEGIS (safety), METIS (progress), NIKE (evidence) and EPISTEMICS (uncertainty), compute GO, DELIBERATE, VERIFY or BLOCK decisions. By default it only observes, logging decisions like WOULD BLOCK without interrupting; enforcement is opt-in. A tool router predicts the smallest useful set of tools, also in observe mode by default. It adds a sidebar to OpenCode's TUI. Credentials are stored in ATHENA's own config with restrictive permissions, live TypeSafe tests are opt-in and regular CI is offline. A demo mode allows offline development.

What the decision model decides

TypeQuestionAnswers
ChoiceShould this agent step proceed?GO · DELIBERATE · VERIFY · BLOCK
ChoiceWhich tools does this step need?(inferred)—

How it works

  1. OpenCode agent proposes a step
  2. Tool router predicts the needed tools
  3. Jev-backed modules assess safety, progress, evidence and uncertainty
  4. Decision recorded; in enforce mode, blocking decisions stop the tool

Strengths

  • Observe-only default for safe rollout
  • Credentials kept out of project files
  • Offline CI and demo mode

Caveats

  • OpenCode only
  • Enforcement is opt-in, so default mode does not block

Getting started

npm install -g athena-jev && athena install && opencode (then athena setup for the key)

At a glance

Type
Project
Status
beta
Language
TypeScript
Runtime
OpenCode plugin + CLI
Has code
yes
Has tests
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • ATHENA gives coding agents reflexes - an open-source cognitive control layer powered by TypeSafe Jev.

    hellogumbo/awesome-jev

Frequently asked questions

Will it block my agent?

Not by default; it observes and logs until you enable enforce mode.

What decisions does it make?

GO, DELIBERATE, VERIFY or BLOCK, plus a smallest useful tool set.

Where is the key stored?

In ATHENA's own config directory with restrictive permissions.

Related projects

  • jev-classifier

    Local gateway and MCP tools that let Jev predict which tool a coding agent should call next, from suggestions to enforced choices.

    MCP serverbetaTypeScript
  • jev-awesome-skills

    Open-source Jev skills for Claude Code, Codex, Cursor and Grok: one client to gate, route, guard, verify, triage, review and act.

    ProjectbetaTypeScript
  • fable-jev

    Pairs Jev with the get-fable lifecycle harness to give coding agents fast tool and skill routing and verbatim-preserving compaction.

    ProjectprototypeTypeScript
  • jevdo

    Strict-allowlist shell harness where Jev walks a TOML decision tree to pick a bash command, flags and paths, confirming writes.

    ProjectprototypePython
  • interlinked-cli

    Local guard layer for coding agents that blocks dangerous tool calls by deterministic rules, with optional Jev judgments.

    CLIbetaTypeScript
  • jev-decision-layer

    MIT plugin and local MCP server that hands an agent's closed questions to Jev or local Laya, with policy gates and receipts.

    Agent skillproductionPython