AI agents call subscribe to retrieve information from Agent Bus without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
The subscribe operation is fundamentally a data retrieval mechanism. It establishes a subscription to a message channel so the agent can read incoming messages. This is equivalent to listening on a queue or subscribing to a data feed—a passive read operation with no side effects on the system, data, or other agents beyond the subscription state itself.
From the tool's definition Tool subscribes agent to a channel to receive messages ('puts a copy in your inbox'). The primary effect is receiving/retrieving data (messages) from a channel—a read operation. No modification, deletion, or execution of code occurs.
Attacks that exploit this kind of access
Subscribe this agent to a named channel. After subscribing, any send_channel(channel) puts a copy in your inbox. It is categorised as a Read tool in the Agent Bus MCP Server, which means it retrieves data without modifying state.
Register the Agent Bus MCP server in PolicyLayer and add a rule for subscribe: 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 Agent Bus. Nothing to install.
subscribe 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 subscribe 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 subscribe. 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.
subscribe is provided by the Agent Bus MCP server (mustaphasteph/agent-bus). 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 →