Recued
Get started

Documentation

Recued holds the space between you and AI. AI reads your world — relationships, decisions, recipes, memory, inbox, calendar, files, CRM — through Recued, and acts on it through Recued, so your data and your accounts' credentials stay local on machines you control. Every action an AI takes is granted in advance, checked at one enforcement boundary, audited, and reversible.

These docs are the practical guide. New here? Start with the Get started funnel — it walks you through installing your server and pairing a client step by step.

Core concepts

  • The control plane. Recued is the layer between you and any AI. AI connects over MCP; it never touches your accounts directly. Reads and actions both pass through Recued, where they are gated, logged, and undoable.
  • Four surfaces. A self-hosted server is the sole orchestrator, identity root, audit authority, and the only thing that touches your accounts. A narrow Browser Bridge actuates the DOM and shows OS notifications. A thin webclient is display and input only. The cloud gives your server a name and a certificate and hosts the marketplace — it routes none of your traffic and never decrypts your data.
  • Recipes and packs. A recipe is a declarative unit of work. A pack bundles recipes (and the ingredients they use) into one installable, grantable unit. Packs are the bundle and publish unit in the marketplace.
  • Grants and the boundary. Nothing runs unless you have granted it. Grants are scoped, every action is recorded, and reversible actions can be undone.

Guides

  • Install the server — one-click VPS, Docker, or a binary. See Get started.
  • Pair a client — connect the webclient to your own server through the cloud-mediated pairing flow.
  • Browse and install — find recipes and packs in the marketplace and install them into your surface.
  • Publish — share your own recipes and packs. See the Publish overview; authoring happens in the Kitchen, presentation and submission in the dashboard.

Plans and policies

Deeper reference

Engineering documentation — architecture, the recipe schema, the decisions log, and contribution guides — lives in the open-source repository on GitHub.