AI agents call refine_search to retrieve information from ContextCore without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Despite the empty description, the tool name and context clearly indicate this is a search/query refinement utility that accesses indexed local files. It performs read-only operations on a search index to return relevant chunks. No evidence of data modification, code execution, deletion, or financial operations. The confidence is slightly reduced due to the missing description, but the context is sufficiently clear.
From the tool's definition Tool is listed as a sibling alongside search and retrieval-focused tools (fetch_content, get_file_content, list_files, get_codebase_context). Server explicitly indexes and retrieves file content using hybrid search.
Documented attack patterns abuse exactly the kind of access refine_search gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and ContextCore, and nothing reaches the server without passing your rules. This is the rule we recommend for refine_search:
{
"version": "1",
"default": "deny",
"tools": {
"refine_search": {}
}
}refine_search is read-only, so it stays allowed — but everything else on the server is denied unless you say otherwise.
Free to start. No card required.
refine_search. It is categorised as a Read tool in the ContextCore MCP Server, which means it retrieves data without modifying state.
Register the ContextCore MCP server in PolicyLayer and add a rule for refine_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 ContextCore. Nothing to install.
refine_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 refine_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 refine_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.
refine_search is provided by the ContextCore MCP server (lucifer-ux/contextcore). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from ContextCore, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
16 ContextCore tools catalogued and risk-classified — across an index of 43,000+ MCP servers.