Ask normally
You use Dvara like a normal LLM chat: research a site, inspect a repo, or prepare a local change.
How it works
AI asks. Dvara governs.
Chat requests become controlled browser, workspace, and patch actions through one local policy layer.
Core loop
- Ask in chat.
- Dvara checks client, grant, domain, and tool policy.
- The managed capability runs locally.
- Result, artifact, and audit entry stay visible.
Simple flow
The product in six steps
Dvara detects local work
Browser, workspace, and patch requests become clear governed actions instead of hidden execution.
Policy checks the request
Client identity, tokens, grants, domains, blocked paths, redaction rules, and emergency stop are checked first.
The managed capability runs
Dvara uses an isolated browser, read-only workspace tools, or a patch artifact path under the desktop runtime.
You keep the write boundary
Agents can propose patches. Humans review and apply selected changes from the desktop UI.
Everything is auditable
Actions, approvals, failures, artifacts, and summaries stay visible in a local audit trail.
What it is