Execution lifecycle operations for executable elements (agents, workflows, pipelines). Supported operations: confirm_operation, approve_cli_permission, execute_agent, complete_execution, continue_execution, abort_execution, prepare_handoff, resume_from_handoff These operations manage runtime exec...
Part of the DollhouseMCP server.
Free to start. No card required.
AI agents invoke mcp_aql_execute to trigger processes or run actions in DollhouseMCP. Execute operations can have side effects beyond the immediate call -- triggering builds, sending notifications, or starting workflows. Rate limits and argument validation are essential to prevent runaway execution.
mcp_aql_execute can trigger processes with real-world consequences. An uncontrolled agent might start dozens of builds, send mass notifications, or kick off expensive compute jobs. PolicyLayer enforces rate limits and validates arguments to keep execution within safe bounds.
Execute tools trigger processes. Rate-limit and validate arguments to prevent unintended side effects.
{
"version": "1",
"default": "deny",
"tools": {
"mcp_aql_execute": {
"limits": [
{
"counter": "mcp_aql_execute_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}See the full DollhouseMCP policy for all 5 tools.
These attack patterns abuse exactly the kind of access mcp_aql_execute gives an agent. Each links to the full case and the policy that stops it:
Other execute tools across the catalogue. The same approach applies to each: rate-limit and validate the arguments.
Execution lifecycle operations for executable elements (agents, workflows, pipelines). Supported operations: confirm_operation, approve_cli_permission, execute_agent, complete_execution, continue_execution, abort_execution, prepare_handoff, resume_from_handoff These operations manage runtime execution state. Unlike CRUD operations (which manage definitions), Execute operations handle the execution lifecycle: - execute_agent: Start a new execution (returns goalId and stateVersion for tracking) - complete_execution: Signal successful completion once the goal is done - continue_execution: Resume a previously paused execution with the same goal parameters - abort_execution: Abort a running execution, rejecting further operations - confirm_operation: Confirm a pending operation that requires user approval (Gatekeeper flow) - approve_cli_permission: Approve a pending CLI tool permission request - prepare_handoff: Serialize goal progress into a portable handoff block for session transfer - resume_from_handoff: Resume agent execution from a handoff block with integrity validation IMPORTANT: Execute operations are potentially destructive (agents can perform any action) and non-idempotent (calling execute_agent twice creates two separate executions). ⚠️ SECURITY: Do not auto-allow this endpoint in your host settings (e.g., Claude Code settings.json). Each execution should require explicit human approval. Auto-allowing bypasses the per-operation confirmation gate. While DangerZone verification and element deny policies still provide protection, the primary human review checkpoint is lost. Canonical loop: 1. Call execute_agent once to start the goal and receive { goalId, stateVersion, activeElements, safetyTier, ... }. 2. After each chunk of work, use mcp_aql_create: { operation: "record_execution_step", ... }. 3. Read record_execution_step.autonomy.continue and any autonomy.notifications to decide whether to continue, pause for a human, or handle a gatekeeper block. 4. When the goal is finished, call complete_execution. Use continue_execution only when an already-started goal was paused and you are resuming it with the same goal parameters. It is not the normal next call after execute_agent. Quick start examples: { operation: "execute_agent", params: { element_name: "code-reviewer", parameters: { objective: "Review code" } } } Next lifecycle step — use mcp_aql_create: { operation: "record_execution_step", params: { element_name: "code-reviewer", stepDescription: "Reviewed auth module", outcome: "success", findings: "Found 2 security issues" } } { operation: "complete_execution", params: { element_name: "code-reviewer", outcome: "success", summary: "Completed review" } } { operation: "abort_execution", params: { element_name: "data-collector", reason: "User requested cancellation" } } { operation: "continue_execution", params: { element_name: "rubric-qa-agent", previousStepResult: "Verified citation set", parameters: { run_dir: "/app/run", deliverable_path: "/app/run/output.docx" } } } { operation: "confirm_operation", params: { operation: "execute_agent" } } { operation: "approve_cli_permission", params: { request_id: "req-123", decision: "allow" } } { operation: "prepare_handoff", params: { element_name: "code-reviewer" } } { operation: "resume_from_handoff", params: { element_name: "code-reviewer", handoff_block: "..." } } Discover required parameters — use mcp_aql_read: { operation: "introspect", params: { query: "operations", name: "execute_agent" } }. It is categorised as a Execute tool in the DollhouseMCP MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Dollhouse MCP server in PolicyLayer and add a rule for mcp_aql_execute: 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 DollhouseMCP. Nothing to install.
mcp_aql_execute 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 mcp_aql_execute 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 mcp_aql_execute. 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.
mcp_aql_execute is provided by the Dollhouse MCP server (@dollhousemcp/mcp-server). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Deterministic rules across all 5 DollhouseMCP tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.
Free to start. No card required.
4,600+ MCP servers and 31,000+ tools scanned and risk-classified.