Access control for AI agents
Grant Auth
Connect your tools once. Use any agent next.
Give a new coding agent the access it needs in under a minute, without handing it provider credentials or repeating setup across every service.
9 connected services1 agent identity<60s target setup
Trusted profileBuild agent
ActiveConnect the stack you already use
GitHubVercelConvexClerkPostHogSearch ConsoleSlackLinearSupabase
One control plane
Access that follows you between agents
Grant separates broad provider authorization from the narrower access each agent receives.
- 01
Connect services
Authorize Grant once. Provider credentials stay encrypted and server-side.
- 02
Choose a boundary
Select read-only, full, or custom provider capabilities and resources.
- 03
Connect any agent
Run one setup command, then ask for provider work in ordinary language.
You stay in control
One place to see, change, and stop access.
Native provider permissionsPer-agent profilesResource boundariesRotation and revocationDecision audit trail
Open beta