Your coding agent finished. Now find the PR, the preview, and the right session.

The agent has fixed a checkout bug but you still need to locate the pull request, review the code changes, test the build, and open the appropriate preview. Remembering which session contains the fix is crucial, as losing it means starting over. An agent session captures decisions, failed attempts, commands, and follow-up questions to avoid rebuilding previous work.

The software Cogpit organizes this workflow, serving as a GUI for Claude Code, OpenAI Codex, and GitHub Copilot CLI. It operates on your machine using existing CLI logins. The source material demonstrates using Cogpit with a checkout bug scenario, including ClickUp task management, output and file editing, session resumption, test inspection, and Vercel preview checking.

Cogpit groups sessions by project, allowing search and comparison across CLIs. It enables inspecting edits by turn or reviewing the full diff. The GitHub panel integrates pull requests, file changes, reviews, and Actions runs. After a successful build, the Vercel panel displays production and preview deployments with accessible build logs.

The Browser panel streams a selected tab from the agent-browser, allowing interaction with the page as the agent does. Named browser sessions maintain separate persistent profiles for project-specific browsing contexts. To try Cogpit, install and sign in to Claude Code, Codex, or GitHub Copilot CLI, then download Cogpit for macOS, Windows, or Linux.

The documentation covers setup for Vercel, GitHub, ClickUp, and browser automation. Feedback on context loss while working with coding agents is appreciated.