Read Working Memory
Start every session with context. Use Context Bundle when you need owner identity, AI Identity, active scope, active rules, and Working Memory together. Use Working Memory alone for the lighter daily briefing.
When to Use
At session start:
- Beginning of a new OpenHands conversation or canvas run
- Returning to a project after a break
- When context about recent work would help
- At the start of a multi-agent canvas node (e.g. planner, coder, reviewer)
During session:
- User asks "what am I working on?" or "what's my context?"
- User references recent priorities or decisions
- Need to understand what's been happening across tools or agent nodes
Skip when:
- Already loaded this session
- User explicitly wants a fresh start
- Working on an isolated, context-independent task
Usage
Via MCP (Preferred in OpenHands)
When Nowledge Mem MCP is configured in OpenHands:
Via CLI
Prefer Context Bundle for startup or multi-agent canvas sessions:
Read Working Memory alone when you only need current priorities:
If it succeeds but reports exists: false, say there is no Working Memory briefing yet.
Multi-Agent & Orchestrator Routing
In OpenHands or SDK orchestrations (such as Raft, Multica, Paseo):
- Set
NMEM_AGENT_ID="<agent-slug>"(e.g.planner,coder,reviewer,architect) before launching the child agent or node to identify the specialist persona. - Set
NMEM_SPACE="<space name>"or pass--space "<space name>"when that run or canvas workspace should scope to a dedicated knowledge lane. - OpenHands
UserPromptSubmithook automatically injects the Context Bundle at session start.
What You'll Find
The Working Memory briefing contains:
- Active Focus Areas — Topics currently engaged with, ranked by recent activity
- Priorities — Items flagged as important or needing attention
- Unresolved Flags — Contradictions, stale information, or items needing verification
- Recent Activity — What changed in knowledge base since the last briefing
- Deep Links — References to specific memories for further exploration
