$ceo-dna
local-first · no hooks · source-available

Your agents inherit
how you build.

ceo-dna parses the transcripts your coding agents already wrote to disk — no hooks, no config changes — and reads your engineering judgment off them: when you abstract, when you stop and ask, what you reject. Not your indentation — your decisions, compiled into a stylepack every agent you run inherits.

how it works

Parse. Private.
Inherit.

No interrogation. No setup ritual. ceo-dna does not ask how you work — it reads it off the decisions already recorded in your transcripts and synthesizes the pattern.

01

Parse

Your agents already wrote it all down.

Every Claude Code session leaves a transcript on disk. ceo-dna parses the ones already there — read-only. It never modifies them, hooks your agent, or changes a single config.

02

Private

Offline by default. Nothing leaves your machine.

Parsing and redaction run entirely locally. The network is touched only when you pass --synthesize with your own API key — and then it goes straight to your provider, never through us. Run ceo-dna inspect to see exactly what was KEPT and what was MASKED.

03

Inherit

Your judgment is now a portable artifact.

Synthesis compiles the decisions you make repeatedly — your abstraction triggers, your stop-and-ask boundaries, what you reject — into a stylepack an agent acts on. Hand it to a new agent. Commit it with your team. Let it evolve as you do.

privacy — first class

You own the lab.
The sample never leaves.

A genetics service ships your DNA to a server you do not control. This is the opposite: the analysis runs on your machine, against files already on your disk. Offline by default — no network unless you pass --synthesize with your own key. The one question a senior engineer asks — what do you actually see? — is answered as commands you can run and verify, not promises.

  • ceo-dna inspect

    Shows exactly what was captured and what was redacted — KEPT vs MASKED — before anything is stored.

  • ceo-dna purge

    Deletes our local capture for a session, date, or range. It never touches your agents' transcripts.

  • ceo-dna pause

    Stops capture immediately, for as long as you want. ceo-dna resume turns it back on.

  • ceo-dna synthesize

    Offline by default. The only command that can touch the network — and only when you pass --synthesize with your own API key. It goes straight to your provider; cli.ceo is never in the path.

ceo-dna inspect 2026-05-19bash
$ ceo-dna inspect 2026-05-19

session 7f3a · claude-code · 41 events
  prompt    "refactor the auth guard to ..."    KEPT
  tool      edit src/auth/guard.ts              KEPT
  decision  abstracted after 3rd duplication    KEPT
  env       OPENAI_API_KEY=[████████████]       MASKED
  path      /Users/[████]/.ssh/id_ed25519       MASKED
  prompt    "deploy with token [████████]"      MASKED

42 KEPT · 3 MASKED · 0 bytes left this machine

Source-available under BUSL-1.1 — so you can read the source, audit the redaction logic, and verify the privacy model yourself. Not a paywall: free for internal and production use. It converts to MPL-2.0 four years after release. Read the license.

Redaction is automatic and pattern-based (secrets, credentials, sensitive paths). No automated redaction is perfect, and it is not a general PII scrubber — which is exactly why inspect shows you everything before it is stored.

the payoff

Meet your
developer DNA.

After synthesis, a local read-only dashboard surfaces the judgment patterns read off your transcripts — concrete, evidenced, and yours. Then generate a shareable DNA card: text-only, no sensitive data, just the patterns.

$ ceo-dna cardshareable · text-only

Your DNA says…

  • You call yourself test-first, but you write tests first only 30% of the time — except on auth, where it is 100%.
  • You extract a helper on the 3rd duplication, not the 2nd. You tolerate the copy in tests.
  • You reject the agent’s first solution ~40% of the time — almost always for over-engineering.
  • You stop and ask before anything destructive; on UI and tests you run autonomous.

What does your developer DNA look like? → npx ceo-dna

Read-only. On localhost. Yours.

ceo-dna dashboard serves your stylepack, activity, and (if you enabled synthesis) spend on a local port. Nothing is published. Nothing phones home.

Run the full walkthrough — start, backfill, synthesize, open the dashboard:

where does this go?

Three distinct problems.
One visible path.

ceo-dna is complete on its own terms. The ladder is here so you can see where it goes — not so you feel pushed. Each tier unlocks a new category of problem, when you have it.

Personal

now

free · local · today

“I want to know and preserve how I work.”

npx ceo-dna gets you here today. Capture, redaction, dashboard, and local synthesis — free, offline, no license key. Your DNA stays on your machine.

Team DNA

later

cloud · when you’re ready

“I want my team to build with consistent style.”

Shared, cross-developer DNA, hosted governance, and cloud sync — your coaching baked in. cli.ceo provides the managed layer on top. The local CLI stays free.

cli.ceo platform

later

governance · when you outgrow local

“I want to govern how my agents operate, org-wide.”

Agent governance, audit trails, policy gates, and managed runners — the full picture, when your team outgrows the local tool.

ceo-dna is where your engineering style lives. cli.ceo is where it scales.

get started

What does your developer
DNA look like?

Ten minutes, entirely on your machine. Run it, see what your stylepack says about you — and tell us if it is wrong.

Star on GitHubRead the READMEsource-available · BUSL-1.1 → MPL-2.0