kris-hansen/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.

What it solves

Comanda provides a terminal-native runtime for managing durable, self-improving agentic workflows within a software repository. It addresses the problem of "hallucinated" completion—where an agent claims a task is done when it isn't—by implementing observable exit criteria and automated quality gates (like tests and linters) that must be passed before a workflow finishes.

How it works

Comanda allows users to describe a desired outcome in English, which it then converts into a YAML-based program. This program defines a sequence of steps, parallel processes, or agentic loops. It coordinates various AI models (Claude, Gemini, OpenAI, local models via Ollama/vLLM) and tools, passing data through files or standard input/output.

Key technical components include:

  • Agentic Loops: Stateful iterations that persist state, refine prompts based on prior results, and run quality gates after each step.
  • Codebase Indexing: Captures repository structure and symbols to create a traversable knowledge graph stored in a SQLite FTS5 database for semantic recall.
  • Durable Memory: A project-local semantic memory system that allows agents to recall specific project decisions, constraints, and facts.
  • MCP Integration: Workflows can be exposed as tools and prompts via the Model Context Protocol (MCP).

Who it’s for

Developers who want to govern and reuse durable agent work in their repositories without switching to a full agent framework or a visual automation platform.

Highlights

  • Deterministic Quality Gates: Use existing tests, security checks, and linters as the definition of "done."
  • Knowledge Graph Visualizer: An interactive local browser view for navigating complex codebase dependencies and architecture.
  • Multi-Agent Coordination: Orchestrate different models (e.g., Claude for review, OpenAI for implementation) in a single pipeline.
  • Checkpointing: Ability to interrupt and resume long-running agentic loops from the last checkpoint.
  • Semantic Memory: Opt-in, namespace-scoped local recall of project-specific facts and decisions.

Related

  • Dispatch
    Command Center: An Agentic Coding Environment for Production Quality

    Command Center is an agentic coding environment designed to reduce the time spent reviewing AI-generated code by providing structured walkthroughs and refactoring agents to ensure production-grade quality.

  • Project
    johnlindquist/mdflow

    A Git-native control plane that turns Markdown files into repeatable, version-controlled AI agent workflows with built-in behavioral evaluations.

  • Project
    AgentEra/Agently

    Agently is a Python framework for building reliable AI services. It provides structured request/response contracts, instant streaming of partial results, observable actions (tool calls), versioned skills, and a signal‑driven workflow engine (TriggerFlow). The library abstracts model providers, supports model pools, and includes built‑in actions for Python, shell, web browsing, SQLite, etc. Designed for production‑grade assistants, internal copilots, and AI‑backed APIs that need debuggability, restart‑safe workflows, and strict output validation.

  • Project
    agentspan-ai/agentspan

    A durable runtime for AI agents that prevents execution loss during process crashes and supports long-running, event-driven workflows with human-in-the-loop approvals.

  • Project
    henryqin1997/statem

    A command-line state machine that turns AI agent workflows into inspectable graphs of states and executable checks to ensure reliability in long-running tasks.