Concepts

How Recued works

Understand the server, Browser Bridge, webclient, cloud, and the enforcement boundary between AI and your accounts.

Recued is a control plane around AI activity. It separates orchestration, actuation, display, and optional internet conveniences so no thin client or cloud service silently becomes the authority over your data.

#The four surfaces

Surface Responsibility Does not own
Server Orchestration, identity, credentials, warehouse, grants, audit Hosted UI or marketplace presentation
Browser Bridge Narrow DOM actions and operating-system notifications Recipes, policy, or long-lived authority
Webclient Display and human input Engine execution or account credentials
Cloud Marketplace, naming, certificates, and coordination conveniences Your warehouse, prompts, or provider credentials

#One enforcement boundary

Every operation reaches the same policy boundary before execution. The boundary combines the installed capability, the actor, the target contract, the granted operation, and the operation's risk tier.

text
AI request
   ↓
resolve actor + contract + capability
   ↓
grant and approval checks
   ↓
local execution
   ↓
audit record and reversible result

Reads are granted explicitly. Writes can require approval. Destructive actions remain approval-gated even when lower-risk work is allowed more broadly.

#Cloud boundary

The cloud may help a server obtain a stable name and certificate and hosts the public marketplace. It does not become a relay for ordinary recipe traffic and does not receive the credentials stored by your server.

#Continue

Read Security and privacy for what crosses each boundary, or see Recipes and packs for how installable capabilities fit it.

Recued is local first; your server remains the authority.

Recued Docs

Search documentation

Start typing to search the documentation.