Return deterministic triage items for tracked work. Surfaces blocked statuses, stale active work, expiring or expired tracked statuses, near-term event staleness, and tracked namespaces missing status or lifecycle structure. Use this instead of broad natural-language search when you explicitly wa...
AI agents call memory_attention to retrieve information from Munin Memory without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
memory_attention retrieves and presents pre-computed or filtered data from the persistent memory store (SQLite database). It has no side effects—it does not create, modify, delete, or execute operations. The tool is purely informational, designed to help surface existing tracked items that need attention. This is a classic Read operation: query-like retrieval with no state changes.
From the tool's definition Tool description states it 'Returns deterministic triage items' and 'Surfaces' tracked work statuses—purely retrieval operations with no modification.
Attacks that exploit this kind of access
Return deterministic triage items for tracked work. Surfaces blocked statuses, stale active work, expiring or expired tracked statuses, near-term event staleness, and tracked namespaces missing status or lifecycle structure. Use this instead of broad natural-language search when you explicitly want what needs attention.\n\nIf this is your first memory operation in this conversation, call memory_orient first. It is categorised as a Read tool in the Munin Memory MCP Server, which means it retrieves data without modifying state.
Register the Munin Memory MCP server in PolicyLayer and add a rule for memory_attention: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Munin Memory. Nothing to install.
memory_attention is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the memory_attention rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for memory_attention. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
memory_attention is provided by the Munin Memory MCP server (magnus-gille/munin-memory). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →