Execute steps from an implementation plan, generating code changes. This tool orchestrates the execution of plan steps, using AI to generate the actual code changes needed for each step. Execution Modes: - single_step: Execute a specific step by number (requires step_number) - all_ready: Execute ...
AI agents invoke execute_plan to trigger actions in Context Engine MCP Server. 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.
This tool executes code generation and plan steps whose side effects are dependent on the plan contents and step configuration. While it generates code changes "by default" in preview mode, the tool is fundamentally designed to execute operations that modify a codebase.
From the tool's definition Tool description states it "Execute[s] steps from an implementation plan, generating code changes" and "orchestrates the execution of plan steps, using AI to generate the actual code changes".
Documented attack patterns abuse exactly the kind of access execute_plan gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and Context Engine MCP Server, and nothing reaches the server without passing your rules. This is the rule we recommend for execute_plan:
{
"version": "1",
"default": "deny",
"tools": {
"execute_plan": {
"limits": [
{
"counter": "execute_plan_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}execute_plan stays usable, but rate-capped — a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
Free to start. No card required.
Execute steps from an implementation plan, generating code changes. This tool orchestrates the execution of plan steps, using AI to generate the actual code changes needed for each step. Execution Modes: - single_step: Execute a specific step by number (requires step_number) - all_ready: Execute all steps whose dependencies are satisfied - full_plan: Execute steps in dependency order (respects max_steps limit) Output: - Generated code changes for each step (preview by default) - Success/failure status for each step - Next steps that are ready to execute - Overall progress tracking You can pass a saved plan_id instead of the full plan JSON. Important: - By default, changes are shown as preview only (apply_changes=false) - Set apply_changes=true to actually write the generated code to files - Use stop_on_failure=true (default) to halt on first error. It is categorised as a Execute tool in the Context Engine MCP Server MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Context Engine MCP Server MCP server in PolicyLayer and add a rule for execute_plan: 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 Context Engine MCP Server. Nothing to install.
execute_plan 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 execute_plan 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 execute_plan. 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.
execute_plan is provided by the Context Engine MCP Server MCP server (kirachon/context-engine). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from Context Engine MCP Server, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
50 Context Engine MCP Server tools catalogued and risk-classified — across an index of 43,000+ MCP servers.