Selection criteria
How we choose CLIs
A useful CLI lets an agent name the resource it needs, read the result and check a change before applying it.
What belongs here
We include tools that work with a repository, service, dataset or product account without driving a browser. We look for explicit resource names or IDs, output a script can parse and a way to run without a prompt.
A dry run, diff, validation result, sandbox or local file makes changes easier to review. Not every tool has all of these. The directory notes describe the parts that matter.
What we leave out
AI agent apps, package managers, interactive-only terminal interfaces, common shell tools and browser automation. They can be useful. This list covers a narrower job.
Unofficial tools are welcome
A community CLI can fit if it is maintained, explains how credentials work and gives scripts readable output. Check its repository and permissions before handing it an account token.
Suggest a tool
Read the full criteria and open a pull request with one tool. Describe what the tool lets an agent do. Leave out popularity claims and vendor slogans.
Who maintains this
This website shows the awesome-ai-cli list from GitHub. Vincent Schmalbach maintains the list and also makes Asana CLI. Its page says so. Entries are selected for their fit with the criteria; inclusion is not a security review.