AI agents call memory_prune_candidates to retrieve information from Memory without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
The tool only retrieves and lists memory entries matching certain criteria (0 hits, not updated within threshold). It performs no writes, deletions, or side effects. The description explicitly states it 'does not delete anything' and is 'for human review only', confirming pure read semantics.
From the tool's definition 'List memory entries' and 'does not delete anything' — explicitly a read-only listing operation for human review
Attacks that exploit this kind of access
List memory entries with 0 hits that have not been updated within the threshold. For human review only, does not delete anything. It is categorised as a Read tool in the Memory MCP Server, which means it retrieves data without modifying state.
Register the Memory MCP server in PolicyLayer and add a rule for memory_prune_candidates: 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 Memory. Nothing to install.
memory_prune_candidates 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_prune_candidates 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_prune_candidates. 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_prune_candidates is provided by the Memory MCP server (joshdougall/memory-mcp). 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 →