OPERON

AI Terminal Cockpit — a fully agentic, multi-provider terminal AI with 185+ tools, persistent memory, browser & desktop automation, and a self-improvement loop.

v3.1.0 · Phase 11 Build · 1,896 tests passing · MIT License

Quick Install

The recommended installer handles everything — Python dependencies and the
Playwright Chromium browser binary (the ~120 MB download that pip install skips).

macOS / Linux

git clone https://github.com/OperonAgent/Operon.git
cd Operon
./install.sh

Windows

git clone https://github.com/OperonAgent/Operon.git
cd Operon
powershell -ExecutionPolicy Bypass -File install.ps1

Any platform (Python)

python install.py            # core + recommended + browser binary
python install.py --full     # also voice, databases, screen capture, etc.

That's it. The installer:

1. Verifies Python ≥ 3.9
2. Creates a .venv virtual environment
3. Installs Operon and its dependencies
4. Downloads the Chromium browser binary (the step everyone forgets)
5. Registers the operon command

Then launch:

operon          # or: python main.py

On first run a setup wizard configures your AI providers and API keys.

What gets installed

Why the browser is a separate step: pip install playwright installs only the
Python package. The actual Chromium browser binary requires a separate
playwright install chromium. Operon's installer (and a runtime self-heal hook)
do this for you — if you ever try to browse without it, Operon downloads it on demand.

Verify your install

operon --check-deps          # dependency status report
operon  →  then type /doctor # full in-app health check

Re-run dependency provisioning anytime:

python -m core.bootstrap            # core + recommended + browser
python -m core.bootstrap --full     # everything
python -m core.bootstrap --browser  # just the Chromium binary
python -m core.bootstrap --check    # status only

Updating

operon --version     # show version + check GitHub for a newer release
operon --update      # pull the latest version (git checkout) + refresh deps

--update is safe: it does a fast-forward git pull and aborts if you have
uncommitted local changes (so it never clobbers your edits). For ZIP/binary
downloads, grab the newest release from the Releases page instead.

Makefile shortcuts

make install        # full install + browser binary
make install-full   # every optional feature
make browser        # just the Chromium binary
make check          # dependency status
make run            # launch Operon
make test           # run the test suite

Pre-built apps (no Python needed)

Each tagged release ships standalone binaries via GitHub Actions:

These bundle Python and all dependencies. (Browser automation in the bundled app
still downloads Chromium on first use.)

Highlights

• 185+ tools — files, shell, git, HTTP, databases, PDF, vision, image gen, TTS
• 8+ AI providers — OpenAI, Anthropic, OpenRouter, Ollama, LM Studio, Jan, and more
• Persistent memory — FTS5 + LanceDB semantic vectors + Obsidian vault sync
• Browser & computer use — Playwright automation + pyautogui desktop control
• SWE agent — automated issue → fix → test → PR loop
• Multi-agent mesh — parallel/pipeline delegation with named specialist roles
• Voice pipeline — speech-to-text, text-to-speech, multimodal
• Self-improvement — synthesizes new Python tools from conversation trajectories
• 60+ slash commands — /kanban, /checkpoint, /mesh, /swe, /voice, …
• Claude Code-style TUI — persistent input bar, token counter, model HUD

Requirements

• Python 3.9+ (3.11+ recommended)
• macOS, Linux, or Windows
• Optional: an Ollama install for fully-offline local models (no API key needed)

Documentation

• Setup Guide — Operon_Setup_Guide.pdf (build: python generate_setup_guide.py)
• Full technical reference — Operon_Documentation.pdf (build: python generate_docs.py)
• Competitive comparison — Operon_Comparison.pdf (build: python generate_comparison.py)
• Build all at once: make docs

License

MIT © 2026 Kunwar Mann