DebugTools Logs and SQL MCP

5.3.0 adds read_target_application_logs and get_last_sql_statements to read recent in-process records through the target Agent's HTTP endpoints. Invoke the target method first, then filter by invocation time or keywords to reduce unrelated records.

Prerequisites

The IDEA project must have an active DebugTools connection, the target must use the 5.3.0 Agent, and its HTTP port must be reachable. Remote or Kubernetes connections also need the corresponding HTTP port forwarding.

Record Scope

The log tool currently captures logs printed by DebugTools itself inside the target JVM. It does not automatically capture application Logback, Log4j, or complete standard output. The SQL tool reads statements that entered the DebugTools SQL printing flow and passed its filters. Enable SQL Printing first.

Read Recent Logs

Tool: read_target_application_logs.

Parameter	Required	Type	Description
projectPath	No	string	Target IDEA project path.
connectionId	Conditional	string	Required when several active connections exist.
limit	No	integer	Default 100, restricted to 1–200.
since	No	integer	Unix timestamp in milliseconds. Keeps records at or after this time; default 0.
level	No	string	Exact log level match, ignoring case, such as ERROR. Other levels are not included automatically.
keyword	No	string	Case-sensitive substring of the log message.

Request example:

{
  "connectionId": "demo-connection",
  "limit": 20,
  "level": "ERROR",
  "keyword": "UserService"
}

Record fields are timestamp, level, logger, thread, message, and throwable. timestamp is a Unix timestamp in milliseconds. throwable is null when no exception is present.

Read Recent SQL

Tool: get_last_sql_statements.

Parameter	Required	Type	Description
projectPath	No	string	Target IDEA project path.
connectionId	Conditional	string	Required when several active connections exist.
limit	No	integer	Default 50, restricted to 1–200.
since	No	integer	Unix timestamp in milliseconds; default 0.
keyword	No	string	Case-sensitive substring of SQL text.

Request example:

{
  "connectionId": "demo-connection",
  "limit": 10,
  "keyword": "user"
}

Record fields are timestamp, sql, consumeTimeMillis, dbType, and applicationName. SQL is stored in the current printing format; execution time is in milliseconds.

Results and Buffer Limits

Normal results from both tools use a structured wrapper with an array of records in data:

{
  "success": true,
  "data": [],
  "error": null,
  "requestId": "example-request-id",
  "timestamp": "2026-10-08T00:00:00Z"
}

The result contains the most recent matching records, ordered from oldest to newest. Each buffer retains at most 500 records. The log buffer also evicts old records based on message size. Both clear when the target process restarts. SQL auto-save is not required, and these tools do not read IDEA's local ~/.debugTools/sql history files.

Result	Meaning and action
success=true, data=[]	No matching records in the current buffer. Check collection scope, SQL printing, and filters.
LOGS_UNAVAILABLE	The log request or result parsing failed. Check Agent version, HTTP port, and connection.
SQL_HISTORY_UNAVAILABLE	The SQL request or result parsing failed. Check Agent version, HTTP port, and SQL capability.
CONNECTION_AMBIGUOUS	The target is ambiguous. Select connectionId from the error's candidates.
NO_CONNECTION	Establish an active connection first.

An unavailable capability is not an empty record set. Even a successful query does not prove that all application logs or all database statements were captured.