discover_tools
Multi-modal tool search engine with 14 scoring strategies: keyword, fuzzy (typo-tolerant), n-gram (partial words), prefix, semantic (synonym expansion), TF-IDF (rare tags score higher), regex, bigram (phrase matching), domain cluster boosting, dense (TF-IDF cosine), neural embedding tool RRF, dom...
This record as markdown: /tools/io-github-homenshum-nodebench/discover-tools.md
What discover_tools does on Nodebench
AI agents call discover_tools to retrieve information from Nodebench without modifying anything. It is typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Why discover_tools is rated Low
discover_tools is fundamentally a search and retrieval function. It queries an index of tools using multiple scoring strategies (keyword, semantic, neural embedding, etc.) and returns ranked results. There is no evidence of data modification, deletion, code execution, or financial transactions. The tool is passive—it searches and ranks existing metadata without side effects. This is a classic Read operation.
From the tool's definition "Multi-modal tool search engine" that "Returns ranked results with quick refs." The tool performs discovery and querying across a tool catalog without modifying, deleting, or executing external operations.
Attacks that exploit this kind of access
The rule that runs discover_tools safely
PolicyLayer is an MCP gateway: it sits between your AI agents and Nodebench, and checks every tool call against a rule you set before the call runs. Nothing changes on the server itself. For discover_tools, this is the rule to start with:
discover_tools is read-only, so it stays allowed. Everything else on the server is denied unless you say otherwise.
The button opens the PolicyLayer dashboard: create your workspace, connect Nodebench, apply this rule, and every discover_tools call is checked against it from then on.
Questions about discover_tools
Multi-modal tool search engine with 14 scoring strategies: keyword, fuzzy (typo-tolerant), n-gram (partial words), prefix, semantic (synonym expansion), TF-IDF (rare tags score higher), regex, bigram (phrase matching), domain cluster boosting, dense (TF-IDF cosine), neural embedding tool RRF, domain RRF (Agent-as-a-Graph bipartite graph), upward traversal (tool→domain→siblings), and execution trace edges (co-occurrence mining). Returns ranked results with quick refs. Use mode=. It is categorised as a Read tool in the Nodebench MCP Server, which means it retrieves data without modifying state.
Register the Nodebench MCP server in PolicyLayer and add a rule for discover_tools: 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 Nodebench. Nothing to install.
discover_tools 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 discover_tools 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 discover_tools. 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.
discover_tools is provided by the Nodebench MCP server (nodebench-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
More on Nodebench, and thousands of servers like it.
This server
Across the catalogue