Skip to content

Get Started

Taskless turns corrections into rules so the same mistake never slows you down twice. You can start using it right now with no sign-up required.

Wondering whether Taskless covers your stack? Supported Languages lists every language and file format a rule can read.

Run the Taskless CLI in any repository that has a coding agent configured (e.g., a .claude/, .cursor/, .opencode/, or .agents/ directory):

Terminal window
npx @taskless/cli

It lists the coding agents it supports, with the ones your project already uses checked, and installs the Taskless skill into each one you pick. It also writes a launcher, ./.taskless/bin/cli.mjs, that runs exactly the CLI version you just installed, and offers to commit what it wrote. Say yes: an agent working in a git worktree only sees what is committed.

Ask your coding agent to onboard the project with Taskless. The installer ends by printing that prompt for you to copy: Onboard this project with Taskless. The agent reads through your codebase, its review comments, and its commit history, finds the conventions worth capturing as deterministic rules, writes every one it can enforce locally, and publishes a short report of what they catch. It asks you one question first.

Ask your coding agent to create a rule using natural language:

“Create a taskless rule that catches direct fetch() calls without error handling”

Taskless is local-first. The agent authors the rule on your machine whenever it reasonably can, then validates it locally, with no account or API calls. You don’t have to pick how the rule gets enforced. Taskless works that out from what the rule depends on. When a rule genuinely needs the Taskless service, the agent tells you before using it. See Authoring Rules for how the destination gets chosen.

Run your rules against your codebase:

Terminal window
./.taskless/bin/cli.mjs check

Any matches are reported with file, line, severity, and message. The command exits with code 1 on error-severity matches, so it’s suitable for CI pipelines.

You’re using Taskless.

  • Create rules: your coding agent generates static analysis rules locally from natural language descriptions
  • Improve rules: iterate on existing rules to fix false positives, broaden coverage, or adjust severity
  • Check rules: run all rules against your codebase from the CLI or CI
  • Delete rules: remove rules you no longer need

With a free Taskless account, you unlock:

  • AI-powered rule generation: Taskless generates higher-quality rules using its API, with validation and test cases
  • GitHub integration: create rules from PR comments, issues, and the dashboard, delivered as pull requests
  • Advanced features: organization controls, rule-based memory management, auditing, and rule tracing

Ready to connect? See Set up Taskless Cloud.

  • Be specific about the pattern. “Detect direct fetch() calls without error handling” beats “improve error handling.”
  • Include code examples. A “bad” example (what should be caught) and a “good” example (what it should look like instead) improve generation quality.
  • Specify the language. Taskless can infer it, but stating it explicitly avoids ambiguity.
  • Explain the “why”. Context about why the pattern is a problem helps generate better messages and fixes.