Scan any codebase in 3s, then verify what your AI builds

Notifications You must be signed in to change notification settings

Fork 0

Star 2

BranchesTags

Open more actions menu

Folders and files

NameName

Last commit message

Last commit date

Latest commit

History

3,113 Commits

3,113 Commits

.ana

.ana

.claude

.claude

.github

.github

.husky

.husky

packages/cli

packages/cli

website

website

.gitignore

.gitignore

.prettierrc.json

.prettierrc.json

AGENTS.md

AGENTS.md

CHANGELOG.md

CHANGELOG.md

CLAUDE.md

CLAUDE.md

CODE_OF_CONDUCT.md

CODE_OF_CONDUCT.md

CONTRIBUTING.md

CONTRIBUTING.md

LICENSE

LICENSE

README.md

README.md

SECURITY.md

SECURITY.md

eslint.config.mjs

eslint.config.mjs

package.json

package.json

pnpm-lock.yaml

pnpm-lock.yaml

pnpm-workspace.yaml

pnpm-workspace.yaml

tsconfig.base.json

tsconfig.base.json

turbo.json

turbo.json

Repository files navigation

Anatomia is the engineering judgment your AI doesn't have. Four agents scope, plan, build, and verify every change. Contracts are sealed before code is written — typed assertions the verifier checks against the code, not Build's account of what it did. Every run produces a proof chain entry — what was asserted, what was found, what shipped. A fifth agent learns from that record and promotes what it finds to rules that shape future builds. Not opinion. Mechanical proof.

Scan any project in 10 seconds

npx anatomia-cli scan .

No install. One command. Here's what you'll see:

┌─────────────────────────────────────────────────────────────────────┐ │ inbox-zero web-app │ │ TypeScript · Next.js · Prisma → PostgreSQL (63 models) │ └─────────────────────────────────────────────────────────────────────┘

Stack ───── Language TypeScript Framework Next.js Database Prisma → PostgreSQL (63 models) Auth Better Auth AI Vercel AI · OpenAI Payments Stripe Testing Vitest UI Tailwind CSS Services Resend · Sentry · PostHog · Upstash (+2 more) Deploy Vercel · GitHub Actions

Surfaces ──────── web Next.js · Vitest api TypeScript · Vitest cli TypeScript

Intelligence ──────────── Activity 7 contributors · 22→18→25→19 weekly Hot files clean.ts (14), ai-categorize.ts (11), schema.prisma (9) Docs README.md · CONTRIBUTING.md · .env.example + 2 more Pre-commit typecheck + lint

Run ana init to scaffold 8 skills (5 core + ai-patterns, data-access, api-patterns)

Install

Like what you see? Install globally to use the ana command directly:

npm install -g anatomia-cli

Requires Node.js 22+. To update: npm update -g anatomia-cli

Quick start

ana init # generate context + agents ana init commit # persist to git (so teammates get it too) ana doctor # verify installation is healthy claude --agent ana # start with "hey Ana" — context loads first claude --agent ana-setup # enrich with your team's knowledge (optional, recommended, ~10 min)

After updating the CLI, run ana init again to refresh scan data and skill detection. Your rules, context, and proof chain are preserved.

init runs scan automatically and works standalone — no Claude Code required. The pipeline and setup require Claude Code.

Tell Ana what you want to build. It'll investigate the codebase, surface tradeoffs, and push back if the approach has problems. When the scope is right, it hands off to Plan, Build, and Verify.

Removing

rm -rf .ana .claude/agents .claude/skills AGENTS.md

If Anatomia created your CLAUDE.md (check git blame), remove that too. If you ran ana init commit, revert that commit: git revert . Your source code is never modified — only infrastructure files.

What it does

Scan + init

ana scan reads your project and detects framework, database, auth, testing, services, conventions, and patterns. In monorepos, scan identifies each surface (package or app) and detects per-surface commands. Re-running ana init refreshes scan data without overwriting your edits.

ana init writes that intelligence to files agents read:

scan.json — full structured scan data for agent consumption

CLAUDE.md and AGENTS.md — cross-tool project context

5 core + 3 conditional skill templates with scan-driven Detected sections

16 stack-specific gotchas with compound triggers

Setup (claude --agent ana-setup) bridges the gap between what scan detects and what your team knows. A ~10 minute session that investigates your codebase, asks 2-3 questions, and writes enriched context. After setup, agents understand your product and decisions — not just your stack.

The pipeline

Stage Agent Role Produces

Think Ana Thinking partner — scope, investigate, advise, push back scope.md

Plan AnaPlan Architect — design + sealed contract spec.md + contract.yaml + plan.md

Build AnaBuild Builder — implement spec, prove it works Code + tests + build_report.md

Verify AnaVerify Fault-finder — reads spec and code, skips Build's report verify_report.md

Learn AnaLearn Proof analyst — runs between cycles Stronger skills and system improvements

Proof intelligence

Every pipeline run writes a proof chain entry — here's one:

┌─────────────────────────────────────────────────────────────────────┐ │ ana proof │ │ Add Stripe Webhooks 2026-04-29 11:07│ └─────────────────────────────────────────────────────────────────────┘

Result: PASS

Contract ──────── 14/14 satisfied · 0 unsatisfied · 0 deviated

Assertions (6 of 14) ────────── ✓ Webhook endpoint verifies Stripe signature before processing ✓ Events are processed idempotently using the Stripe event ID ✓ Failed signature check returns 400, not 500 ✓ Unrecognized event types return 200 without processing ✓ Migration adds idempotency_key column with unique constraint ✓ Existing checkout and billing portal flows pass without modification

Findings ──────── [risk · scope] Signature verification uses direct string comparison — timing-safe equality not enforced [debt · scope] No retry mechanism for failed event processing — transient DB errors will drop events silently [observation · monitor] Webhook handler is 340 lines with a switch that will grow with every new event type

Timing ────── Total 52 min Think 4 min Plan 15 min Build 22 min Verify 11 min

Each entry adds to a proof chain. ana proof health tracks the trajectory across runs — first-pass verification rate, risks per run, hot spots where findings cluster, and what to fix next. When patterns recur, proof promote turns them into skill rules that reach the next build. proof audit groups active findings by file. proof stale flags findings whose files changed since discovery.

Commands

Scan and init

Command Description

ana scan [path] Detect stack, conventions, patterns. --quick for surface-only, --json for structured output

ana init Generate .ana/ context and .claude/ agent definitions. Re-run after CLI updates to refresh

ana init commit Commit infrastructure files to the artifact branch

ana doctor Check project health and configuration. --json for CI

ana config show Display current ana.json settings

ana config delete Remove a config field, reverting to detected default

Pipeline

Command Description

ana work start Start a work item, record timestamp

ana work status Show pipeline state for active work

ana work complete Archive plan, write proof chain entry

ana artifact save Save pipeline artifact with hash verification

ana artifact save-all Save all artifacts in a plan directory atomically

ana verify pre-check Run contract seal verification

ana pr create Create PR from verified build

Proof intelligence

Command Description

ana proof Display proof chain entry

ana proof health Quality trajectory dashboard

ana proof audit Active findings grouped by file

ana proof close Close resolved findings with reason

ana proof promote Promote findings to skill rules

ana proof strengthen Commit skill edits and close findings

| ana proof audit --new | Filter to findings since last learn session | | ana proof audit --since | Filter to findings after ISO date | | ana proof stale | Show findings with staleness signals | | ana proof context | Query proof chain for file context |

Learn

Command Description

ana learn end Mark session boundary — next Learn session knows what's new

Setup

Command Description

ana setup Enrich context with team knowledge (Claude Code agent)

ana setup check Validate context file quality

ana setup complete Validate context and finalize setup

ana agents Agent dashboard — list deployed agent definitions

ana agents model [agent] [model] Show or set agent model overrides

Works with

Built for Claude Code. The pipeline, agents, and skills are Claude Code native.

Scan output (AGENTS.md, CLAUDE.md) works with any AI tool that reads markdown.

Development

git clone https://github.com/anatomia-dev/anatomia.git cd anatomia && pnpm install && pnpm build cd packages/cli && pnpm vitest run

See CONTRIBUTING.md for extension guides and ARCHITECTURE.md for the module map.

This project is built with Anatomia. The .ana/ directory is the proof — every feature was scoped, planned, built, and verified through the same pipeline this tool installs for you.

License

MIT

About

Verified AI development. Ship with proof.

anatomia.dev

Topics

cli

typescript

ai

verification

developer-tools

claude-code

context-engineering

proof-chain

Resources

Readme

License

MIT license

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy

Uh oh!

There was an error while loading. Please reload this page.

Activity

Custom properties

Stars

2 stars

Watchers

0 watching

Forks

0 forks

Report repository

Releases 7

v1.1.3

Latest

May 22, 2026

+ 6 releases

Packages 0

Uh oh!

There was an error while loading. Please reload this page.

Uh oh!

There was an error while loading. Please reload this page.

Contributors

Uh oh!

There was an error while loading. Please reload this page.

Languages

TypeScript 93.9%

MDX 2.9%

CSS 2.8%

Other 0.4%