Live Beta — free to use during beta. Paid tiers are marked "Free during beta" until monetization launches.
Live Beta — free to use during beta. Paid tiers are marked "Free during beta" until monetization launches.
One interface. Every AI. It flies itself.
Connect the AI agents, APIs, local models, and command-line tools you already use — and run them together from a single browser tab. You bring your own subscriptions and API keys; AI Cockpit never resells model usage. Open source under the Apache License 2.0 (licence and terms).
Then let it fly. State a goal and Autopilot Cruise plans it, runs it as a durable workflow, and hands back the result — you clear the one critical moment. Underneath, AutoCognition gives your agents a shared local brain: they see, message, and call each other, vote on answers, and route work to the best-fit model. Prefer to fly by hand? Press Enter and it dispatches to the agent you picked, exactly as before — Cruise takes over only when you ask for it with the flight PLAN button.
Free forever with your own models. One click on Stay Free and the agent cards run real open-source coding agents — Aider, Goose, opencode — on your local Ollama models: no key, no subscription, works offline. Bring your Claude/Codex subscription only if you want the premium agents. The Full edition registers the install once with a free account — no card, ever — which also opens the community Marketplace; Stand-alone (V1) skips even that and makes no network calls at all. Optional paid tiers add Autopilot Cruise, AutoCognition, and remote control (Everywhere) — never bigger quotas on the local features.
One product, three editions — one codebase; the edition just sets what's switched on: Full (accounts + Marketplace + relay), Stand-alone (V1) (everything local, zero network), and the hosted Web Demo.
A picture's worth a thousand words — try the live demo right in your browser: real UI, real themes, zero install. It's a hosted showcase — features are described, not run — so download the app to use them for real, on your own machine.
Download Free
Explore Features
Marketplace
Documentation
AI Models
language Everywhere Relay
81 downloads
The difference in one line: you state a goal, Cockpit does the rest — and it survives a crash or a closed tab, because the whole run is durable.
Type a goal in plain language — "summarise these three PDFs, then draft a reply."
A planner breaks the goal into steps and picks the best agent or model for each.
Steps execute as a durable chain — checkpointed, resumable, auto-recovering if interrupted.
The result lands back in your chat. Clear the one moment that needs you; skip the busywork.
Sequential, parallel, or a vote between models — all on one durable engine. See how orchestration works →
A local-first web application that connects all your AI agents and tools into one unified workspace. Install it on your own Linux server once — then access it from any browser on your local network.
Configurable slots for a paid CLI (Claude, Codex, or Qwen Code), a free-tier REST provider (Gemini or Groq, with a free API key), and a local LLM server (Ollama, LocalAI, or Dual — both at once) — plus cloud Workers (Groq, OpenRouter) and local CLI Tools (ffmpeg, whisper, pandoc), all in one interface.
Let Autopilot Cruise plan and run a goal end-to-end, or drive by hand: Single, Pipeline (A→B→C), Broadcast (parallel), or a model vote. AutoCognition gives agents a shared bus to message, call, and route work between each other — all on one durable, resumable engine.
Runs on Apache + PHP + Bash. No WebSocket daemon, no Node.js server, no persistent process. Survives Apache restarts. Your data never leaves your machine unless you choose it.
Attach any file. Text and code is inlined. PDFs route to pandoc. Audio routes to whisper.cpp. Images route to vision. Everything ends up in your prompt intelligently.
Build a searchable library of code snippets and knowledge items. Keyword search is instant and offline. Semantic deep search uses your local Ollama embeddings model.
Dark, light, and mixed modes. 18 Protanopia-safe Dalton variants designed for red-green color deficiency. Auto-contrast safety net ensures readability on every theme.
Five supplier cards across the top. Click to select, expand for full controls. Each shows real-time usage metrics from your actual log files.
Every card shows real usage read from your own log files — not estimates.
No credit card required for these providers. Add an API key and start using models with context windows far beyond any free tier's limits.
Free, open, local-first. Runs on Linux, ARM Linux, macOS, and Windows (via WSL2) — every other device connects through the Everywhere relay instead.
Get AI Cockpit — Free