AI agents call speak to retrieve information from Voice MCP without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
The speak tool is a unidirectional output operation that retrieves/presents information (text) in audio form. It has no capability to modify, delete, execute, or move data. The worst-case misuse would be speaking unexpected or sensitive information aloud, which is a read/disclosure risk rather than a destructive or execute-level risk.
From the tool's definition Tool description states 'Speak text aloud to the user using text-to-speech' — this outputs information to the user via audio without modifying any data, executing code, or producing side effects beyond audio playback.
Attacks that exploit this kind of access
Speak text aloud to the user using text-to-speech. It is categorised as a Read tool in the Voice MCP MCP Server, which means it retrieves data without modifying state.
Register the Voice MCP server in PolicyLayer and add a rule for speak: 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 Voice MCP. Nothing to install.
speak 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 speak 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 speak. 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.
speak is provided by the Voice MCP server (jochiang/voice-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 →