Twill

About Twill

Twill scores how ready a GitHub repository is for AI coding agents and then does something about the gaps. The free scorecard is the entry point: paste a public repo URL, or sign in to analyse a private one, and it grades the codebase against OpenAI's Agentic Legibility framework and the Assistants and Agents Build Hour guidance — the conventions that determine whether an agent dropped into a repo can find the build command, understand the module boundaries, run the tests and avoid the traps a new human hire would need a week to learn. Scoring runs inside a hosted shell via the Responses API rather than by static pattern-matching, so it reflects what an agent actually encounters. The score is useful on its own as a diagnostic, and it is the kind of artefact a team can put in front of an engineering lead to justify the work. Beyond the scorecard, Twill sells credit-metered runs against your codebase with Claude Code, Codex CLI and OpenCode as the recommended agents, aimed at clearing backlog while improving the repo's legibility at the same time. Enterprise adds a codebase audit that identifies automation opportunities, bring-your-own API keys and private-cloud deployment inside your VPC. Open source projects get it free under a fair-use policy.

Key Features

Tags