Tell AI once. Taskless turns your team's corrections into deterministic rules for AI coding agents. Correct once. Enforced forever.
npx @taskless/cliYou describe what you want enforced, and Taskless decides how to check it on every commit, PR, IDE session, and agent run. Each rule compiles to a real check with its own pass/fail fixtures, and no model runs during enforcement. The CLI is open source and works without signing up.
Markdown instructions like AGENTS.md, skills/, and .cursor/rules are injected into prompts and followed inconsistently. Taskless uses analysis instead: a syntax tree for code (ast-grep), the text itself for prose (Vale), your existing linter, or a cross-file runtime check.
- The right check for each convention. 25+ languages and prose, run locally, handed to your linter's own format when possible.
- Deterministic. Rules give the same result on every run. They aren't an AI reviewer's opinion.
- Fixtures included. Every rule proves it catches what you intended before it touches your code.
- One source of truth. Rules come from team decisions, not scattered markdown files you have to keep in sync.
- Gets cleaner over time. Corrections accumulate into rules, so later pull requests need fewer of them.
In a repository with a coding agent configured (a .claude/, .cursor/, .opencode/, or .agents/ directory), run the installer. It detects your agent tools and installs the Taskless skill into each.
npx @taskless/cliAsk your agent to help you get started with Taskless. It reviews your codebase and history and helps turn the strongest candidates into rules. Or describe one yourself:
Create a taskless rule that catches direct fetch() calls without error handling.
Then check your code, locally or in CI:
./.taskless/bin/cli.mjs checkMatches are reported with file, line, severity, and message, and the command exits 1 on error-severity matches.
For conventions bigger than one team. Free for individual developers and small teams.
- Provenance. Every rule traces back to the PR or decision it came from.
- Governance. One rule applies across every developer and agent in your org.
- Authenticity. Rule code is signed to prevent tampering.
- GitHub integration. Tag
@tasklesson a PR or issue to turn a correction into a rule. - AI-powered rule generation, with validation and test cases.
- Website: taskless.io
- Documentation: docs.taskless.io
- Blog: taskless.io/blog
- Taskless Cloud: taskless.io/cloud
- Sign in: app.taskless.io
- npm: @taskless/cli
- Contact: hello@taskless.io