← All agentsGet started free

AI Code Agent APIs

Discover and run AI coding agents via API. Code review, PR descriptions, commit messages, README generation and more — one API key, pay per run.

AI code agents automate the repetitive parts of software development. Instead of switching context to an AI chat tool, you call an API endpoint from your CI pipeline, git hook, or editor and get structured output back. Every agent on Runcept accepts a standard POST request and returns JSON — no prompt engineering, no session management, no model selection.

Common use cases

  • Automated PR descriptions on every pull request via GitHub Actions
  • Pre-commit hooks that suggest conventional commit messages from staged diffs
  • CI pipeline step that blocks merges if code review score is below threshold
  • README generation for internal tools that never get documented
  • Batch code review across multiple repos before a compliance audit

Text Diff

Compares two texts and returns the differences — line-by-line, word-by-word, or as a unified patch. Powered by the jsdiff library.

Free

SQL Formatter

Formats SQL queries with sql-formatter — supports PostgreSQL, MySQL, SQLite, BigQuery, Snowflake and more. Free and open source.

Free

FAQ

Do I need to set up any AI models or API keys?

No. Each agent on Runcept handles its own model selection and prompting. You just call the endpoint with your input and get output back. Your only credential is your Runcept API key.

Can I use these agents in GitHub Actions?

Yes. All Runcept agents are standard HTTP endpoints. Add your RUNCEPT_API_KEY as a repository secret and call POST /api/v1/run from any workflow step.

What happens if an agent fails mid-run?

Credits are automatically refunded on any failed run. You are never charged for an agent that does not return a successful response.

How long do code agents take to run?

Most code agents complete in 5-30 seconds depending on the size of the input. Runcept uses an async job pattern — your request returns a job_id immediately and you poll for the result.