Get started with Recued
Run your own Recued server, pair a client, and install a foundation pack. Your data and your accounts' credentials stay on machines you control; every action an AI takes is granted in advance, audited, and reversible.
Why Recued
Your data stays local
AI reads and acts on your world through Recued. Your accounts' credentials and your data live on your own server — the cloud never decrypts them.
One audited boundary
Every action passes through a single enforcement point: granted in advance, logged, and reversible. Nothing runs that you didn't allow.
Any AI, over MCP
Connect Claude, GPT, or a local model over MCP. Recued is the layer between you and whichever AI you choose.
Free and self-hosted
The whole engine runs on your server at no cost. AI resolves through free-pool keys or your own BYOK — Recued bears no inference bill.
Built to reach anything
Recued is a substrate, not a fixed feature list — it reaches whatever you wire into it, and every connection runs through the same audited boundary.
Connect to any API or service
Reach any HTTP API, plus your mail, calendar, files, and CRM through the connection layer. Credentials are entered once on your server and never leave it.
Wrap any command-line tool
Turn a CLI — ffmpeg, pandoc, OCR, PDF tools — into a server-side step your recipes and AI can call. The tool runs on your machine, never in the cloud.
Your world as one warehouse
Mail, calendar, files, contacts, CRM, and a memory graph in one local store the AI reads through Recued — with provenance and a timeline behind every fact.
Reversible PII aliases
AI works over aliased stand-ins; the real emails, names, and URLs are restored locally on the way back, so the model never sees them.
Public reception, review-first
Accept inbound requests — webhooks, a public endpoint — into a review-then-approve inbox. Nothing acts until you approve it, with the arguments in front of you.
Scheduled and reactive
Run recipes on a schedule, on a trigger — a new email, a CRM change — or as idle-time housekeeping. All on your server, no cloud cron.
Install your server
Pick the path that fits you. Each one ends with a running Recued server.
One-click VPS
Deploy a Recued server to a low-cost VPS from a provider's marketplace template — no terminal required. Pick a provider, click deploy, and your server boots ready to pair.
Best for an always-on server you don't manage by hand.
Docker
Run the self-updating image on any machine with Docker:
docker run -d --name recued -v recued-data:/data -p 8787:8787 ghcr.io/recued/recued:managedThe launcher keeps the server updated from the signed release manifest.
Binary
Download the signed binary for your platform, verify its signature, and run it. The server self-updates from its signed release channel:
./recued serve --data ./recued-dataBest for a Mac, Linux box, or Raspberry Pi you keep on.
Then pair a client
Open the webclient at app.recued.com and pair it to your new server with the code it shows on first boot. The webclient is display and input only — your server runs the work.
Start with a foundation pack
Foundation packs bootstrap your warehouse with useful recipes from day one. Install one, then browse the rest.
Personal Organizer Foundation
Composes your built-in work entities — tasks, notes, and projects — into a day-one organizer.
Install →Personal CRM Foundation
Time-anchored alerts before each meeting, grounded in patterns from your inbox and calendar.
Install →Solo Command Center
A daily cockpit: today's schedule, open commitments, stalled projects, and recent notes.
Install →