Application data. Agent work. Actions.
Scroll to follow the data
Illustrative system · Select a highlighted recordScroll to explore

Infrastructure
for agents
that act.

For companies building agents across business applications. One system for the data they use, the work they produce, and the actions they request.

Application
data.

Give agents the data they need. Reads retain the returned information, its source, and observation time. Background synchronization maintains selected datasets between tasks.

Queryable
state.

Maintain current records and related context across applications. Agents and product interfaces query the same views.

Authorized
actions.

Record a request, check authorization, and retain execution attempts and outcomes. Your product can determine what changed and what remains uncertain.

Retained
context.

Keep application data, generated work, and human corrections connected. Later tasks retrieve prior decisions and their supporting information, even as agents and models change.

Build with
Groundhog.

Start with the data your agents already request. Add synchronization, maintained views, and managed execution as your product develops.

Designed for SDK, API, and MCP integration.

For agents working
across applications.

In development

Start with
the event log.

The durable event log and HTTP API are available now. Synchronization, projections, and managed execution describe the broader product architecture.

About this site.

Groundhog is infrastructure for companies building agents. This site explains the product architecture and the current event log implementation.

The interactive examples use illustrative data. They do not connect to applications or execute external actions. Early access remains a placeholder in this preview.

Read the API reference