Analysis updated 2026-05-18
Define a reusable AI agent, such as a code reviewer, as a Markdown file that any teammate can run.
Build a repeating workflow where one agent implements a change and another agent reviews it until it is approved.
Run project specific agents directly from Claude Code, Cursor, Codex, or another supported coding tool.
Validate an agent definition file before relying on it in a real workflow.
| baldaworks/callee | sliverkiss/mimocode2api | tj/go-gracefully | |
|---|---|---|---|
| Stars | 65 | 65 | 66 |
| Language | Go | Go | Go |
| Last pushed | — | — | 2014-12-27 |
| Maintenance | — | — | Dormant |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Node.js with npm for the setup command, or Go if installing from source, also needs a separate ACP provider for the AI model.
Callee is a command line tool that lets a repository define reusable AI agents and step by step workflows using plain Markdown or YAML files instead of code. Each agent has a clear boundary: it takes some input, may update a shared state object, and returns either a finished result or a structured outcome describing what happened during the run. There are three kinds of agents. A Role is a single agent backed by an AI model. A Sequential agent runs a series of steps in order. A Loop agent repeats a set of steps, such as having one agent implement something and another review it, until a condition like approval is met. Callee itself has no server and does not store long running threads, it works entirely through the command line and relies on a separate project called Norma Runtime to talk to the underlying AI provider. Callee installs two skills into coding tools you already use, such as Codex, Claude Code, Grok Build, Copilot CLI, OpenCode, or Cursor. One skill finds and runs an existing agent or workflow, then reports back the result along with details about how it ran. The other skill helps you write a new agent definition, using a built in template when helpful, and checks that the file is valid before saving it. Setting up a host takes one line typed into a terminal, and does not require installing anything globally first. Once set up, you can describe a task in plain language inside your coding tool, and Callee figures out whether to run an existing agent or workflow, or create a new one. It also has direct command line commands for installing it globally, listing available agents, viewing or validating an agent file, and running an agent or workflow with a message describing the task. Running an agent requires an interactive terminal session because it may need to ask for input or permission along the way. The full README is longer than what was shown.
Callee is a CLI tool for defining reusable AI agents and multi step workflows in Markdown or YAML, runnable through coding tools like Claude Code, Cursor, or Codex.
Mainly Go. The stack also includes Go, Markdown, YAML.
MIT license, use freely in personal or commercial projects.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.