AI agents call classify to retrieve information from Devon without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
The classify tool reads data from a DEVONthink record and returns classification proposals via built-in AI. This is a read-only operation that analyzes content and returns suggestions for categorization, with no side effects, data modification, or destructive capability. The low severity reflects minimal blast radius if misused—an AI could classify documents incorrectly, but no data is altered or deleted.
From the tool's definition Tool provides 'classification proposals' for records, which is a query/analysis operation that retrieves AI-generated suggestions without modifying the record itself.
Attacks that exploit this kind of access
Get classification proposals for a DEVONthink record. It is categorised as a Read tool in the Devon MCP Server, which means it retrieves data without modifying state.
Register the Devon MCP server in PolicyLayer and add a rule for classify: 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 Devon. Nothing to install.
classify 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 classify 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 classify. 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.
classify is provided by the Devon MCP server (mnott/devon). 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 →