Tracking what Claude Code did in your development workflow
2 min read
When you run multiple coding sessions in parallel, keeping track of what actually happened across each codebase gets difficult. This guide explains how to maintain a clear record of your prompts, summaries, touched files, and tasks so you always know the current state of your work.
Understanding the Need for Local Agent Logs
Coding agents work fast, generating code changes, running tests, and modifying project files in rapid succession. When you step away from your Mac or switch between several terminals, retracing those steps requires digging through terminal scrollback or git histories.
A dedicated task memory system solves this by automatically capturing what a tool like Claude Code did during your active sessions. Instead of relying on manual notes or guessing which files were modified, you can inspect a structured log that organizes your workflow by date and time.
What a Session Log Records
To be genuinely useful, a local log needs to capture the concrete details of your agent interactions without adding manual overhead. When you track what Claude Code did, the record should include:
- Your original prompts and instructions
- Summaries of the work completed during the session
- Specific files touched or modified during the run
- Outstanding tasks and pending items
This data should be grouped by day with exact dates and times, making it straightforward to review an entire day of development or pinpoint when a specific change was introduced.
Using Task Memory in Tidepane
In Tidepane, this functionality is handled by a built-in task memory feature. It is on by default and records only while it is running, meaning it never reads anything from before you started the app. All recorded information stays entirely on your Mac.
To open the log, you can use the header control or press Shift-Command-H. Once open, you are not limited to what the agent generated. You can add tasks of your own to the list and mark completed work as verified once you have reviewed the changes. Because Tidepane also lets you view live agent status in the sidebar and interact with built-in chats, having a centralized history complements your active terminal work.
Keeping Your Workflow Transparent
Maintaining visibility into agent actions helps prevent regressions and keeps your team aligned on what was changed in a repository. By reviewing prompts, summaries, and touched files in one persistent view, you spend less time reconstructing context and more time building.
Take the next step by opening your task memory log with Shift-Command-H during your next coding session to verify your agent's completed tasks and add your own items.