Search across both Claude Code memories and AgentDB entries using semantic vector similarity. Returns merged, deduplicated results from all namespaces. Use when native Read/Write is wrong because you need (a) cross-session retrieval by semantic similarity (vector embeddings) not by file path, (b)...
AI agents call memory_search_unified to retrieve information from Claude Flow without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
This is a search and retrieval tool that queries existing data using semantic vector similarity. It has no side effects—it does not create, modify, delete, or execute operations. The blast radius of misuse is limited to unauthorized information disclosure from existing memories and database entries.
From the tool's definition Tool performs 'Search across both Claude Code memories and AgentDB entries' and 'Returns merged, deduplicated results from all namespaces.' The description explicitly states it is for 'cross-session retrieval' and does not modify, delete, or execute…
Attacks that exploit this kind of access
Search across both Claude Code memories and AgentDB entries using semantic vector similarity. Returns merged, deduplicated results from all namespaces. Use when native Read/Write is wrong because you need (a) cross-session retrieval by semantic similarity (vector embeddings) not by file path, (b) namespacing across projects without managing directory layout, or (c) the .swarm/memory.db audit trail. For one-shot file I/O, native Read/Write is fine. It is categorised as a Read tool in the Claude Flow MCP Server, which means it retrieves data without modifying state.
Register the Claude Flow MCP server in PolicyLayer and add a rule for memory_search_unified: 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 Claude Flow. Nothing to install.
memory_search_unified 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_search_unified 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_search_unified. 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_search_unified is provided by the Claude Flow MCP server (claude-flow). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.