DebugTools MCP Workflow

5.3.0 provides status, search, method invocation, HotSwap, logs, and SQL tools that can be combined to verify results after editing code. HotSwap tools currently report submission only. Confirm compilation and reloading in IDEA before using an invocation result as evidence that new code took effect.

Identify the Project and Target

1. Call get_debug_tools_status. Provide projectPath when several projects are open.
2. Select an active connectionId. If none exists, select an explicit PID from list_attachable_jvms and attach, or select and start an exact run configuration.
3. Wait for the connection. For Spring Bean calls, also confirm /spring/ready. Resolve HTTP connectivity issues first.
4. Before hot reload, confirm IDEA Java Debugger is connected to the same JVM. Provide sessionName when several sessions exist.

The DebugTools Agent connection handles method invocation. The Java Debugger connection handles HotSwap and breakpoints. They are separate connections.

Locate the Method and Prepare Arguments

If the class and method are known, call generate_method_args_template directly. If only the HTTP path is known, use search_http_url to locate the Controller, then confirm the method signature and overloaded parameter types.

Fill the template's content values and preserve declaration order. To reuse a saved pre/post script, get its name from list_method_around_scripts, inspect it with get_method_around_script if needed, and pass the name as methodAroundName.

Verify After Hot Reload

1. Call compile_and_reload_modified_files for an explicit Java debugger session.
2. Retain operationId. get_hotswap_operation currently returns a request record without confirming final completion.
3. Confirm successful compilation and class replacement in IDEA's HotSwap UI or notifications.
4. Call the target method with invoke_java_method, selecting resultView as needed.
5. Query records after the invocation using read_target_application_logs or get_last_sql_statements. Provide the target connection explicitly and narrow the result with since and keywords.

Assess result fetch failure separately from method execution failure. Logs cover DebugTools records only; SQL covers statements printed by DebugTools and not filtered out. An empty array does not mean the complete business flow produced no logs or SQL.

Use run_and_invoke

When no active connection exists, run_and_invoke can explicitly start an exact run configuration or attach a specified PID, then submit hot reload and invoke a method. See Hotswap MCP for parameters and return fields.

Use it when there is one clear target and immediate invocation after submitting hot reload is acceptable. It does not wait for HotSwap completion or automatically verify that the Java debugger session and method connection refer to the same JVM.

For multiple connections or strict verification of new code, use the separate steps above. Orchestrated log/SQL queries read the first active connection without filtering by invocation time, so they are not direct evidence for this invocation on a specified connection.

Recover from Failure

New query tools usually return failures with error.code, hint, availableOptions, retryable, and nextAction. Method invocation still primarily uses success and throwable. HotSwap uses errorCode, message, and candidate session lists.

Select an ID or session again if the target is ambiguous. Reconnect if inactive. List scripts again if one is missing. Resolve compilation errors reported by IDEA first. Do not treat submission success, empty records, or unknown capabilities as a passed verification.