openclaw-logfire

What This Skill Does

The openclaw-logfire skill provides comprehensive observability for the OpenClaw agent ecosystem by integrating Pydantic Logfire. This plugin automatically instruments agent workflows using OpenTelemetry (OTEL) semantic conventions specifically tailored for GenAI applications. By capturing the full agent lifecycle, it allows developers to visualize complex interaction chains, monitor token usage, and track tool execution performance. The plugin creates a detailed span tree for every invocation, distinguishing between root agent processes and nested tool calls like file system operations, shell commands, or API requests. It acts as a bridge between your local OpenClaw environment and Logfire’s dashboard, ensuring that every significant step—from initial input to final response—is logged with context on latency and resource consumption.

Installation

To integrate this observability suite, first install the package via the OpenClaw plugin manager: openclaw plugins install @ultrathink-solutions/openclaw-logfire

After installation, you must authorize the plugin by setting the LOGFIRE_TOKEN environment variable in your terminal session or .env file. Finally, enable the integration within your openclaw.json configuration file by setting enabled to true under the openclaw-logfire entry. Once saved, restart your OpenClaw agent instance to initiate the tracing stream.

Use Cases

This skill is essential for teams moving from experimental agent workflows to production-grade deployments. Common use cases include debugging "looping" issues where an agent fails to complete a task, auditing token usage to optimize cost, and identifying which specific tool call is creating a latency bottleneck. It is particularly useful for distributed systems where an agent might trigger actions across different microservices, providing a single source of truth for debugging failures.

Example Prompts

1. "OpenClaw, please run the audit on the server logs and log every tool interaction to Logfire for performance analysis."
2. "I need to track how many input and output tokens were consumed during my last batch processing task; check the Logfire metrics."
3. "Trace the execution path of my recent file management task to see if the write command is being delayed by the file system permissions."

Tips & Limitations

To maintain security, the plugin defaults to redactSecrets: true, which is highly recommended for production environments. Be mindful that while metadata and spans are captured, actual message content and tool outputs are disabled by default for privacy. If you need deep debugging, you can enable these in the configuration, but do so with caution regarding sensitive data handling. Note that all tracing data is streamed directly to Logfire, meaning there is no local persistence; if your network connection drops, those trace spans will be lost.