Use cases
For agents working
across applications.
Current data, related context, and a record of the actions agents take.
Read. Review. Act.
- ReadCRM and billing
- ReviewHuman correction
- ActRequested change
Account agents
Resolve customer
requests.
An agent uses account data, recent activity, and conversation history to resolve a request. It can update a preference and retain the information that informed the change.
Later tasks use previous actions and human corrections as context.
Explore action executionNotificationsMonthly
SourceCRM
Operations agents
Work across
applications.
An operations agent combines account records, project activity, and messages from different applications. Groundhog retains each source and observation time, and maintains views the agent can query.
Explore maintained viewsCustomer activity
One view across applications.
Agent development
Learn from
reviewed work.
Human reviews connect corrections to an agent's work and the data it consumed. Developers use recorded situations and reviewed outcomes to build evaluations. The history remains available when agents or models change.
Explore retained historyInformation returned to the agent.
Generated output and its inputs.
Corrections linked to the output.
Build into
your product.
Shared data, maintained views, authorized actions, and retained history across tasks.
These workflows describe the product architecture. The durable event log is available today.
Read the architecture