Describe a task. CLIO reads your code, makes changes, runs tests, and commits - iterating until the work is done. Interrupt it mid-task to redirect, and it picks up exactly where you left off next session.
Free and open source. ~50MB. Zero external dependencies. macOS and Linux.
CLIO doesn't hand you a snippet and walk away. It reads your code, makes changes, runs tests, fixes errors, and commits - autonomously, from start to finish.
Press any key to stop CLIO mid-task. It pauses, asks what you need, and adapts its approach - like tapping your pair programmer on the shoulder.
Long-term memory stores project discoveries, solved problems, and code patterns. Every new session starts with what CLIO learned last time - no re-explaining.
Run /profile build and CLIO analyzes your session history to learn how you communicate, what you prefer, and how to work with you - across every project.
Spawn multiple AI agents to tackle different parts of a problem simultaneously. File and git locking prevent conflicts. Live output streams in separate tmux, screen, or Zellij panes. Puppeteer mode orchestrates agents across multiple projects in a monorepo or multi-repo setup.
SSH into any machine, run an AI task, get results back. Execute across your entire fleet in parallel - diagnostics, builds, analysis, deployment.
Track token usage, premium request counts, and GitHub Copilot quota with /usage (or /billing - same thing). See your billing rate multiplier, quota remaining, and reset date at a glance.
Built-in secret redaction catches API keys, tokens, passwords, and PII before they reach any AI provider. Your credentials never leave your machine.
Pure Perl. ~50MB at startup. No npm, no pip, no CPAN. Runs on anything from a ClockworkPi uConsole to an M4 Mac. Install via Homebrew, Docker, or a single script.
GitHub Copilot, OpenAI, Google Gemini, DeepSeek, OpenRouter, MiniMax, llama.cpp, LM Studio, SAM, and any OpenAI-compatible endpoint. Switch providers mid-session.
CLIO reads, writes, searches, and refactors files across your entire project. Apply surgical patches across multiple files. Undo any AI change with a single command. Run shell commands, build tools, test suites - all from the same session.
Semantic search understands code structure. Ask "where is authentication implemented?" and CLIO finds it - even if the file doesn't contain that word.
Sessions persist. Resume any conversation with full history intact. CLIO tracks multi-step tasks with structured todos, maintains momentum across long sessions, and exports complete session transcripts to HTML for documentation and review.
Context trimming keeps long sessions sharp - CLIO automatically compresses older messages into summaries so it never loses the thread of what you're working on.
CLIO implements The Unbroken Method - a systematic approach to AI collaboration that maintains continuous context, demands complete ownership, and learns from failure.
Collaboration checkpoints keep the agent aligned. Long-term memory persists what it learns. Todo tracking ensures nothing gets dropped. CLIO doesn't just respond - it works through problems end-to-end with a methodology that scales to real projects.
Start with GitHub Copilot - no API key setup, just log in and go. Swap providers anytime.