Full-text search across papers in the local library
AI agents call search_library to retrieve information from arXiv MCP Server without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
search_library is a read-only operation that searches and returns information from an existing corpus. It has no side effects, does not modify data, execute code, delete records, or move resources. The blast radius of misuse is minimal—an agent could at worst retrieve irrelevant papers or perform many searches, neither of which causes harm.
From the tool's definition Tool performs 'Full-text search across papers in the local library' — a query operation that retrieves data without modification, deletion, or execution of external processes.
Attacks that exploit this kind of access
Full-text search across papers in the local library. It is categorised as a Read tool in the arXiv MCP Server MCP Server, which means it retrieves data without modifying state.
Register the arXiv MCP Server MCP server in PolicyLayer and add a rule for search_library: 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 arXiv MCP Server. Nothing to install.
search_library 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_library 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_library. 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_library is provided by the arXiv MCP Server MCP server (r-uben/arxiv-mcp-server). 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 →