Latitude has launched an open-source platform designed to monitor AI agents in production, allowing users to track agent performance once they interact with real customers. This platform enhances the developer workflow by integrating an MCP server, which provides coding agents direct access to production traces, signals, and datasets, streamlining the process of identifying and addressing issues. The system aggregates thousands of live conversations, enabling teams to analyze user interactions and pinpoint failures, effectively treating agent conversations as vital internal records of customer feedback. The release reflects an increasing focus on agent reliability as businesses scale their AI deployments.

Latitude: Latitude is the team and open-source project building a platform for monitoring AI agents in production. It surfaces agent behavior with real users, identifies recurring failures through clustered traces and signals, and routes insights back into developer workflows via an MCP server integration. The release treats agent conversation data as a primary record of product performance and reliability.
Claude Code: Claude Code is Anthropic’s coding-focused interface for the Claude model, available in terminal, desktop, and IDE environments. Latitude integrates directly with it to capture full multi-turn session transcripts as traces for analysis and cost tracking within the monitoring platform.
latitude-dev: latitude-dev is the GitHub organization that maintains the Latitude open-source project under an MIT license. It oversees the codebase, documentation, and community contributions for the AI agent observability platform.

Developer Workflow: An MCP server allows coding agents to access production traces, signals, and datasets directly, closing the loop between monitoring and code fixes without switching tools.
AI Agent Monitoring: Latitude positions agent conversations as the richest internal record of customer interactions and product issues for teams deploying AI at scale.