Use case · Customer-Facing Agents
Build secure agent-powered products
Safely connect your users’ data, tools, and workflows to AI agents — at scale.
The problem
Your product’s agents act for your users — each with their own accounts, consent, and boundaries. Building that infrastructure is a product in itself.
01
Agents act for many users at once
02
OAuth scopes are too coarse for per-user boundaries
03
Identity breaks across user → app → agent → tool
04
One crossed tenant is a disclosure, not a bug
How your product acts for each user
Provision
Your product creates the user
Each customer gets their own identity in Fabriq through the API.
Connect
They link their accounts
OAuth runs in their name — their credentials, their consent, their grants.
Delegate
The agent acts as that user
Every call is scoped to exactly one customer's connections and permissions.
Isolate
Tenants stay separate
Each request is checked against its tenant and user — nothing pools.
See it
Capabilities
What you get
01
Per-user identity on every action.
Tools always know exactly which user the agent represents right now.
02
Access scoped to each user’s own accounts.
Agents act within what each user connected and granted — nothing pooled.
03
Tenant isolation, built in.
Every request is checked against its tenant and user — with dedicated per-user instances when isolation matters most.
04
Identity verified at every hop.
User → app → agent → tool — each step in the chain is checked, not assumed.
05
A Python SDK for the plumbing.
Provision users, run OAuth in their name, mint scoped tokens — through the API.
06
Consent your users can see and revoke.
Each user connects their own accounts — and disconnecting kills that access immediately.
The payoff
Every agent action is tied back to the exact user it represents.
Questions
Common questions
Related solutions