Security

Private by architecture

Most agent products bolt privacy onto shared infrastructure. Uvy starts from isolation: an isolated environment per customer, with your credentials, your memory, and your audit trail inside it.

One environment per customer

Your Uvy runs in a dedicated, isolated environment. No shared queues, no multi-tenant processing, no co-mingled data. When we say your Uvy, we mean the machine too.

Credentials stay home

Tool credentials are stored and used only inside your environment, scoped to the minimum each workflow needs. Revoke any credential and that capability stops instantly.

Approval gates

Approval-first by default: irreversible and outward-facing actions (sending, publishing, merging, charging) wait for your explicit go in Slack. The gate list is yours to configure.

Your data trains no one

What Uvy learns about your team stays on your environment and improves only your Uvy. Nothing you do is pooled into a shared model or another customer's agent.

Full audit trail

Every action Uvy takes is logged on your environment: what it read, what it did, which model handled which step, who approved what. Export anytime.

Clean offboarding

Leaving means your environment is destroyed: credentials, memory, logs, everything. We keep what's needed for invoicing and nothing else.

The environment

What runs where

A plain-language map of the architecture, because security pages that hide behind diagrams usually have something to hide.

Your Slack workspace
└─ @uvy (app) · The only thing your team sees
Your isolated environment (dedicated container)
├─ Orchestrator · Plans tasks, enforces approval gates
├─ Model router · Sends each step to the right model
├─ Credential vault · Scoped tokens, never leave the environment
├─ Team memory · Voice, formats, preferences
└─ Audit log · Every read, write, and approval
Model providers
└─ Receive only the minimum context per step, under no-training terms
Compliance

In progress, in the open

SOC 2 Type II is in progress. Until the report is in hand we won't put a badge on this page; we'll show you the architecture, the audit trail, and the controls instead.

Working with your security team

Bring your questionnaire

We complete security reviews as part of Fleet onboarding: questionnaires, architecture walkthroughs, and custom data-handling terms. If your team wants the environment inside your own cloud account, ask us about that.

Now onboarding design partners

Your next hire is already in Slack

We set up your isolated environment and wire your first routines in the first working session. Bring real work.

Request a working sessionSee pricing

Or write to [email protected]