A CARL conversation
CARL — Code Automation Readiness Level
How agent-ready is this repo, really?
- How agent-ready is this repo, really?
- What's the smallest change to move from level 1 to level 2?
- Which workflows can I trust an agent to drive end-to-end?
- Where does my codebase still need a human in the loop?
CARL is a free-to-read, proprietary standard for measuring how ready a codebase is for AI-driven workflows. It scores a repository against eight pillars of agent-readiness — from documented invariants to deterministic builds — and resolves to a single level on a 0–6 scale. The point isn't a vanity badge. The point is a shared vocabulary for the moment a human and an agent meet a codebase together.
The spec is published, self-attested, and adversarial-friendly: you publish your level, you publish your evidence, and the ecosystem decides whether to trust it. The scanner here is the reference implementation.
The scale
Seven levels of readiness, 0 to 6
Levels 0–4 measure how ready a codebase is for an agent today. Levels 5–6 are aspirational — beyond today's AI — pointing at a future where the agent creates and then runs a business.
Manual
Scripted
Tested
Reviewed
Agent-driven
Creates a business
Beyond today's AI
Runs a business
Beyond today's AI
Read the full CARL v1.0 specification — the levels, the eight pillars, and the methodology. Free to read; proprietary, all rights reserved.
What CARL offers
The standard is free. The tools are how we keep the lights on.
The CARL spec and methodology are always free to read — see the license for the current posture (proprietary, all rights reserved). The Scanner and the advisory are separate offerings, each priced on its own.
01
The CARL standard
Free
The specification and methodology — the 0–6 readiness levels, the eight pillars, and the self-attestation method. This is the standard itself.
Read the spec02
CARL Scanner
Contact us
Point the reference scanner at a public GitHub repository and get an evidence-backed readiness read against the eight pillars.
Scan a public repo03
Drift-aware advisory
Contact us
Continuous, Claude-powered drift monitoring that tells you which pillar slipped and the smallest change to climb the next level. Onboard through the /builder funnel.
Request accessSee the full pricing breakdown for what each tier includes.
We run it on ourselves
The standard isn't a slide deck. It's how we build.
The monorepo that builds and ships this site is itself assessed against CARL, and its autonomous build agents run at the low end of the ladder today — advisory and human-supervised, roughly Level 0–1. We publish where we actually are rather than claim a level we haven't earned. That's the whole point of the standard: a level is only worth anything next to its evidence.