comanda
The CLI-native orchestrator for AI agent workflows. Run Claude Code, Codex, Gemini CLI & Kimi Code from declarative YAML. Because the terminal is where real wor
What is comanda?
The CLI-native orchestrator for AI agent workflows. Run Claude Code, Codex, Gemini CLI & Kimi Code from declarative YAML. Because the terminal is where real work happens.
Overview of comanda
What it is
Comanda is a terminal-native runtime for orchestrating AI agent workflows. It allows users to describe a workflow in English, inspect the generated YAML program, and run coding agents such as Claude Code, Codex, Gemini CLI, and Kimi Code from declarative YAML files. Workflows are durable, stateful, and can be version-controlled within a repository.
Key features
- Generate and improve workflows from natural-language feedback.
- Run agentic loops with state persistence, checkpointing, and automated quality gates (syntax, security, custom commands).
- Coordinate multiple agents (Claude Code, Gemini CLI, OpenAI Codex, Kimi Code, API models, local models) in parallel or sequential workflows.
- Keep agent work inside the repository model using git worktrees, codebase indexes, explicit path boundaries, and quality gates.
- Expose workflows as MCP tools, HTTP endpoints, or OpenAI-compatible server routes.
- Durable semantic memory with SQLite FTS5 for bounded, project-local recall, and knowledge graph extraction from codebase indexes.
Use cases
- Implementing a feature until tests and security checks pass, with automated loops and quality gates.
- Parallel review of code changes by multiple agents (e.g., architecture review by Claude Code, implementation review by OpenAI Codex, synthesis by Gemini CLI).
- Running agent workflows from the terminal, CI pipelines, or as part of shell pipelines.
- Building and evolving governed, reviewable agent programs that are committed to a repository.
Getting started
bash# macOSbrew install kris-hansen/comanda/comanda# Gogo install github.com/kris-hansen/comanda@latestSee GitHub Releases for prebuilt binaries for macOS, Linux, and Windows.
License and pricing
The project is licensed under MIT and is open source.
Based on the project's public README and repository metadata.