搜索知识库。不指定 collection_name 则搜索全部，返回按 collection 分组的 dict。
AI agents call search to retrieve information from RAG MCP Server without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
The search function queries a RAG knowledge base and returns structured results. This is a read-only retrieval operation with no side effects. It does not create, modify, delete, or execute anything—it simply retrieves and ranks existing data. The low severity reflects minimal risk even if misused; malicious search queries cannot harm the knowledge base or external systems.
From the tool's definition Tool name is 'search' and description states '搜索知识库' (search knowledge base), which retrieves/queries data from vector embeddings and returns grouped results. No modification, deletion, or execution of external operations occurs.
Attacks that exploit this kind of access
搜索知识库。不指定 collection_name 则搜索全部，返回按 collection 分组的 dict。. It is categorised as a Read tool in the RAG MCP Server MCP Server, which means it retrieves data without modifying state.
Register the RAG MCP Server MCP server in PolicyLayer and add a rule for search: 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 RAG MCP Server. Nothing to install.
search 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 search 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 search. 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.
search is provided by the RAG MCP Server MCP server (sglwsjxh/rag-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 →