Will ATHENA start blocking my tool calls unprompted?

No, not by default. enforcementMode is observe, which records decisions such as WOULD BLOCK without interrupting OpenCode, and the tool router also defaults to observe so all host tools remain visible. You opt into enforcement with athena config set cognition.enforcement enforce and an active router with athena config set toolRouter.mode active.

Is there any evidence the judgments are any good?

Not in the README. It describes the four controllers, the four verdict strings and the routing order from coding host to adapter to router to LLM to cognitive runtime, and points at the architecture and tool-router docs, but offers no measurement and no example. Treat it as a scaffold whose verdicts you would need to evaluate on your own traffic.

What leaves my machine, and where does the credential live?

Only compact redacted state goes to Jev; chain-of-thought, raw tool arguments, raw outputs and credentials are not persisted. The TypeSafe and Jev credential is stored by ATHENA under $XDG_CONFIG_HOME/athena, or ~/.config/athena, with restrictive permissions, and is not written into OpenCode configuration, project files, shell exports or telemetry.

What does it take to get running?

Node 20 or newer, a global npm install of athena-jev, then athena install to register with a supported OpenCode version. athena setup stores and validates the TypeSafe credential, and athena install --demo is the offline path. athena doctor diagnoses installation issues, which matters because the integration is global and per-version.

What are the four cognitive controls?

AEGIS for safety, METIS for progress, NIKE for evidence, and EPISTEMICS for uncertainty; together they compute GO, DELIBERATE, VERIFY, and BLOCK.

How do I turn enforcement on?

ATHENA is passive by default with enforcementMode observe; switch with athena config set cognition.enforcement enforce, and enable the tool router with athena config set toolRouter.mode active.

What does the Tool Router do?

It predicts the smallest useful capability set, and it defaults to observe so all host tools stay visible; switching it to active is a separate opt-in.

CommunityJev Agents & MCP

Athena Jev

OpenCode extension putting a Jev-backed judgment layer between a coding agent and its tools. Four controllers, AEGIS, METIS, NIKE and EPISTEMICS, emit GO, DELIBERATE, VERIFY or BLOCK, and a tool router predicts the smallest useful capability set. Observe mode is the default.

See highlights

Open github.com

Stars: 2
Language: TypeScript
License: MIT
Published by: Community
Author: MartinesEmanuel
Last verified: Oct 2026

MartinesEmanuel/athena-jev

User repository on GitHub

View on GitHub

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

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

ATHENA is an independent open-source cognitive control layer for coding agents, powered by TypeSafe Jev. It adds native OpenCode observability while keeping the agent in control by default. athena install registers ATHENA globally for supported OpenCode versions. On first use, run athena setup to store and validate a TypeSafe/Jev credential in ATHENA-owned configuration. For offline development, use athena install --demo.

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import

Works forJev for Classification Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Four named controllers, AEGIS for safety, METIS for progress, NIKE for evidence and EPISTEMICS for uncertainty, resolve to four verdicts.
enforcementMode defaults to observe, recording verdicts such as WOULD BLOCK without interrupting OpenCode; enforce is a config change.
The tool router predicts the smallest useful capability set and also defaults to observe, so every host tool stays visible until set active.
Credentials live in ATHENA-owned config with restrictive permissions, never in OpenCode config, project files, shell exports or telemetry.

Watch out

The README is a landing page. It names four controllers and four verdicts and links to docs/architecture.md and docs/tool-router.md, but publishes no benchmark, no accuracy figure and no worked example of what makes METIS return DELIBERATE or NIKE demand VERIFY, so there is no way to judge the judgments from what is written. Installation mutates your OpenCode setup globally via npm and athena install; athena repair recreates ATHENA-owned integration files and athena uninstall removes them, with the details in SECURITY.md. OpenCode V2 gets the full sidebar while V1 is supported with reduced UI capability. Being powered by Jev means a hosted TypeSafe credential configured through athena setup; the offline path is athena install --demo, which is also why regular CI stays offline with live TypeSafe tests opt-in. Node 20 or newer is required. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

JevLoop

Runnable agent harness that moves pick, score and yes-no loop choices to Jev or Laya in milliseconds while reserving the language model for writing, compiling DECISION.md into executable questions.

Jev Agents & MCP#community#jevbest#catalog-import

Community

zjunlp/JevLoopTypeScript

patdown

Node CLI that lints a file tree against fuzzy natural-language rules in one Markdown file, judging each matched file with a swappable TypeSafe-backed judge.

Jev Agents & MCP#community#jevbest#catalog-import

Community

tyler-dot-earth/patdownTypeScript

jev-cli

Command-line toolkit that wraps Jev judgments into scriptable verify, screen, classify, extract, find, rerank, match, route, ask, compact, and batch commands.

Jev Agents & MCP#community#jevbest#catalog-import

Community

Nasrallah-AL/jev-cliTypeScript

jevwire

Wires Jev into agents as seven MCP tools, an embeddable JevDecisionModel library with run functions, and Claude Code hooks at tool and turn boundaries.

Jev Agents & MCP#community#jevbest#catalog-import

Community

Brainwires/jevwireTypeScript

jev-recipes

TypeScript recipe library with 248 focused functions that call Jev for routing, reranking, gating and labelling, returning structured ready-or-review decisions for Node backends.

Jev Agents & MCP#community#awesomejev#catalog-import

Community