Given a user message in context, returns the top-K canonical memories the host should consider injecting into the system prompt or message prefix before LLM inference. This is a recommendation, not a mutation — the host decides whether to use the suggestions. Routes through canonical_index fast-p...
AI agents call memory.suggest to retrieve information from Lore Context without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
memory.suggest retrieves and ranks existing memory data based on similarity to a user message. It has no capability to create, modify, delete, or execute operations. The explicit note that it is 'a recommendation, not a mutation' and that 'the host decides whether to use the suggestions' confirms read-only semantics. This is a search/retrieval tool with negligible blast radius if misused by an AI agent.
From the tool's definition Tool description states: 'returns the top-K canonical memories...This is a recommendation, not a mutation — the host decides whether to use the suggestions.' Performs search/retrieval operations ('Routes through canonical_index fast-path first; falls through…
Attacks that exploit this kind of access
Given a user message in context, returns the top-K canonical memories the host should consider injecting into the system prompt or message prefix before LLM inference. This is a recommendation, not a mutation — the host decides whether to use the suggestions. Routes through canonical_index fast-path first; falls through to 7-stream hybrid search if no canonical hit. Token budget is enforced via tiktoken cl100k_base approximation (±10% drift for non-GPT models). It is categorised as a Read tool in the Lore Context MCP Server, which means it retrieves data without modifying state.
Register the Lore Context MCP server in PolicyLayer and add a rule for memory.suggest: 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 Lore Context. Nothing to install.
memory.suggest 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.suggest 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.suggest. 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.suggest is provided by the Lore Context MCP server (Lore-Context/lore-context). 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 →