Function init_logging 

pub fn init_logging(verbose: bool, log_format: Option<LogFormat>) -> Result<()>

Initializes logging infrastructure.

Sets up tracing with appropriate log levels based on verbosity flag. Writes log messages to stderr, with any embedded URL’s credentials/query string redacted (via a wrapping Write adapter around the fmt layer’s writer) — this covers rmcp and any other dependency’s log lines, not just this project’s own, since a dependency’s tracing output cannot go through this crate’s Debug/escape_error_text redaction paths.

log_format selects text or JSON output. When None (the flag was not passed), LogFormat::resolve consults the MCP_EXECUTION_LOG_FORMAT environment variable; an unset or invalid environment value falls back to text with a WARN-level log line (never echoing the rejected raw value — see the project’s log-injection hardening for this switch).

Both branches are passed through cap_rmcp_log_level, which caps rmcp’s own tracing targets at info regardless of the base filter – see that function’s doc comment for why.

Arguments

• verbose - If true, sets log level to DEBUG; otherwise uses INFO or environment variable override via RUST_LOG
• log_format - --log-format flag value, or None to consult MCP_EXECUTION_LOG_FORMAT

Errors

This function cannot fail—it always returns Ok(()). Multiple calls in the same process will panic rather than returning an error, but this is not a recoverable condition and indicates a programming error.

Examples

use mcp_execution_cli::runner;

// `no_run`: this installs a process-global tracing subscriber, which
// panics if called more than once in the same process.
runner::init_logging(false, None)?;