Give agents durable task state
Agents need project goals, tool results, files, decisions, and unresolved questions to persist beyond a single prompt or run.
Agent workspaces
Durable workspace context for agents that keep working over time.
This page is about Recall Core as a workspace layer for agents: bounded context, explicit tools, resumable task state, and auditable handoff when work spans multiple runs.
Re:Yi official sitePublic content, local custody.Recall Core data stays outside the online service by default.
Keep tool access explicit
A workspace should make it clear which sources, MCP clients, APIs, and documents an agent can use before it starts acting.
Support handoff and restart
When an agent fails, stops, or hands work to another tool, Recall Core should preserve enough context to resume without guessing.
Make automation auditable
Long-running agent work needs visible boundaries, source trails, retry behavior, and clean shutdown paths for real teams.
Agent path
Agents need a workspace, not a longer prompt.
The goal is to let agents work with durable context while users and teams can still understand what was connected, what changed, what failed, and how to stop or recover the run.
- 01Create a bounded workspace
- 02Attach trusted context
- 03Run agents with visible permissions
- 04Audit handoff, retry, and shutdown
Different from AI memory
This is the agent-centered workspace path.
AI memory focuses on the user's long-term context across tools. Agent workspaces focus on bounded, inspectable task environments for systems that act on the user's behalf.