AI agents invoke swap_server to trigger actions in Llauncher. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
Swapping a server model involves stopping a running inference server and starting a new one, which constitutes triggering external operations with side effects. It is not purely destructive (no data is permanently deleted) nor purely a write (it executes process lifecycle changes).
From the tool's definition "Replace the model on this port with a different one" — stops an existing server instance and starts a new one with a different model, triggering external lifecycle operations
Attacks that exploit this kind of access
Replace the model on this port with a different one. It is categorised as a Execute tool in the Llauncher MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Llauncher MCP server in PolicyLayer and add a rule for swap_server: 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 Llauncher. Nothing to install.
swap_server is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the swap_server 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 swap_server. 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.
swap_server is provided by the Llauncher MCP server (shanevcantwell/llauncher). 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 →