An AI agent, quietly running on your machine.
Ravel remembers everything that matters, connects to 118+ tools, and acts on your behalf — running locally, on an open agent network.
An AI workspace you actually run
This is the real Ravel interface — try it right here. Switch views, filter tasks, search memory, press ⌘K.
- ·read 38 messages from memory tree
- ·cross-referenced 12 calendar events
- ▸drafting summary…
Watch Ravel learn in real time
New repositories appear on the open network — Ravel ingests each one, maps it, and adds it to memory. Click any repo to open it.
How Ravel turns your tools into memory
Four stages, running locally — from raw data to autonomous action.
Connects and reads
118+ OAuth integrations stream your data in. Auto-fetch refreshes every 20 minutes so context is never stale.
Builds the memory tree
A local LLM compresses everything into a deterministic memory tree in SQLite — TokenJuice cuts token use up to 80%.
Picks the right model
Each task is routed to a reasoning, fast, or vision model — the cheapest one that can do the job well.
Executes on your behalf
Native tools — web, scraper, coder, voice — let Ravel finish work in the background, then report back.
From install to autonomous agent
Three steps. No configuration, no ramp-up period.
Connect your tools
One-click OAuth into 118+ apps — email, calendar, notes, code, and chat.
Build a memory tree
Ravel compresses everything into a local, deterministic memory you fully own.
Let the agent act
Ravel works in the background, routes tasks to the right model, and gets things done.
Everything a personal agent should be
Memory, integrations, and a full toolbelt — running locally.
1B-token memory
A persistent memory tree stored locally on your machine — no ramp-up, ever.
118+ integrations
One-click OAuth connections to the tools you already use every day.
TokenJuice compression
Up to 80% fewer tokens via smart HTML-to-markdown conversion and summarization.
Native toolbelt
Web search, scraper, a full coder toolset, and voice — built in.
Private by default
A local LLM handles low-level work so your data stays off the cloud.
Smart model routing
Reasoning, fast, or vision — the right model for each task, one runtime.
One agent, every kind of work
Ravel adapts to how you work — here is what it does for different people.
Developers
Ravel reads your repos, tracks PRs, and drafts code with full context of your codebase — no copy-pasting.
Founders
Inbox, calendar, docs, and Stripe in one memory. Ask for a weekly summary and Ravel just knows.
Researchers
Feed Ravel papers and notes; it builds a searchable knowledge tree and surfaces connections you missed.
Always indexing, always learning
Ravel watches the open network and folds every new repository into its memory — here is the stream, live.
Built different on purpose
Most AI assistants run in someone else's cloud. Ravel does not.
Ravel is free — and stays yours
No subscription, no paywall, no per-seat pricing. Ravel is open-source and local-first, so every feature runs on your own machine at no cost.
- ✓All 118+ integrations included
- ✓Full 1B-token memory tree
- ✓Smart model routing and native tools
- ✓Open-source — yours to inspect and extend
Questions, answered
Yes. Ravel is local-first — your memory tree lives in SQLite on your machine. A local LLM handles low-level work, so your data never has to leave your device.
Where Ravel is headed
Integrations
118+ one-click OAuth connections and the auto-fetch engine.
Agent toolbelt
Native web, scraper, coder, and voice tools shipping now.
Desktop presence
The mascot joins calls, lip-syncs, and reacts in real time.
Install Ravel in one line
Free, open-source, and local-first. One command and the agent is yours.
curl -fsSL ravel.sh | sh