_
Athena scans your repo and writes honest, evidence-backed project knowledge that Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI and other coding agents actually read. No guessing, no leaked secrets, and nothing leaves your machine.
Or run instantly:npx project-athena init(Node.js ≥ 22.12)
“Turn any codebase into clear knowledge.” — MIT Licensed CLIBest-in-class context.
See how Athena turns raw repository files into evidence-backed, honest intelligence that AI coding agents rely on instead of guessing.
model User { id String @id email String @unique }
AST Entity Parser
Resolved models, fields & relations without guessing
get_database_schemaathena mcpEvery fact shows its receipts.
AI coding agents fail when they guess. Athena eliminates guesswork by labeling every piece of generated knowledge with its exact certainty tier and source code evidence.
Declared in configuration (such as package.json) or asserted by the developer.
Found through heuristic detection, file signatures, or AST parsing.
Derived from patterns or generated via optional AI providers.
Athena cannot verify the detail with evidence, so it refuses to guess.
// FACT: Declared in configuration or asserted by the developer
"node": ">=22.12" [FACT: package.json (engines)]
"target": "ES2022" [FACT: tsconfig.json:4]Athena says “unknown” instead of guessing. There are no fake confidence scores, no synthesized vanity metrics, and no simulated activity.
What you get
Structured, modular Markdown files stored directly in your repository. Generated files refresh on re-analysis, while developer-owned files like rules.md and custom manual sections are preserved forever.
# Project
## Overview
**Name:** shop
**Project purpose:** UNKNOWN — no description found in manifests.
## Technologies
| Technology | Status | Evidence |
| --- | --- | --- |
| Next.js | DETECTED (high) | package.json |
| Prisma | DETECTED (high) | package.json, prisma/schema.prisma |Works with your agents
Configure integrations in one step. Athena generates target instruction files, registers activity hooks, and provides an on-demand MCP server.
Claude Code
Anthropic's terminal agent.
Cursor
AI-first code editor.
Codex
OpenAI's coding agent (CLI, IDE extension and app).
GitHub Copilot
GitHub's coding agent in VS Code, the Copilot CLI and the cloud agent.
Gemini CLI
Google's open-source terminal agent.
Antigravity
Advanced agentic coding environment.
Windsurf
Codeium's AI editor (now Devin Desktop).
Cline
Open-source coding agent for VS Code and JetBrains.
AGENTS.md
Universal standard for any AI agent.
Any tool or agent supporting AGENTS.md or standard stdio MCP can consume Athena out of the box.
Built for precision and developer control
Stays in sync
Detects repository drift and proposes knowledge updates. Changes are never applied automatically without your explicit review.
Context engine
Show exactly which knowledge documents and sections an agent should read for a specific task or feature, and why.
MCP server
Model Context Protocol stdio server exposing 12 specialized tools so agents query schemas, routes, and rules on demand.
Local web UI
Listens on 127.0.0.1 (port 7432 by default) with token in the URL fragment. Features a knowledge editor, rules manager, sync review, and activity monitor.
Security scan with your own tools
Dispatches native ecosystem security checkers (npm audit, pip-audit, govulncheck, cargo audit, composer audit) without external telemetry.
Diff review & rules checklist
Checks current git diff for facts worth reviewing (secrets, dependencies, schema/route changes) and lists rules from rules.md to verify.
CI checks & pre-commit hook
Run athena sync --check in CI and block commits when knowledge is out of date. Uses the GitHub Action Sabari-Vasan-SM/athena@v0 to review pull requests.
Runs 100% on your machine.
Your proprietary architecture and repository logic remain exclusively on your device.
127.0.0.1 only
Athena listens on 127.0.0.1, port 7432 by default (or the next free port if 7432 is busy). Access token is in the URL fragment (after #), so browsers never send it in requests or Referer headers.
Secrets redacted
Secrets are detected and redacted. Environment variable names are mapped, but values are never stored.
No account
No telemetry account, no cloud login, and no centralized database. Runs 100% locally on your machine.
AI optional
Analysis, documentation, the web UI and the MCP server run fully offline. Two optional commands use the network, and only when you run them: athena security runs your ecosystem's audit tools (such as npm audit), which query their own advisory databases, and athena ai enrich sends redacted documents to the AI provider you configured.
Initialize your project in one command.
Stop letting AI coding assistants hallucinate your architecture. Give them persistent, evidence-backed knowledge in .athena/ right now.