I Built a 'Time Machine' for AI Coding Sessions — Here's Why
Source: Dev.to
The Problem
Last Tuesday I spent two hours in a Claude Code session refactoring an authentication module. Around the 45‑minute mark the AI suggested an elegant token‑rotation approach I’d never considered. It was one of those moments where you think, “I need to remember this.”
I didn’t. By the time I looked, the terminal had scrolled past it. The conversation was a wall of text with no structure. The reasoning behind the change, the back‑and‑forth that shaped it, the intermediate code states – all gone, dissolved into the void of my scrollback buffer.
If this sounds familiar, keep reading.
Why AI coding assistants feel incomplete
AI coding assistants have changed how I work:
- Claude Code – backend logic
- Cursor – frontend iteration
- Gemini CLI – quick prototyping
The productivity gains are real, but there’s a glaring hole in the workflow that nobody seems to address:
AI coding sessions are completely ephemeral.
Think about it. Every other creative tool has history, undo, and review built in:
| Tool type | History feature |
|---|---|
| Video editors | Timeline |
| Design tools | Version history |
| DAWs | Session recall |
| AI coding assistants | None – just a scrolling chat window |
Git tracks what changed, but not why. A commit message like “refactor auth module” does not capture the fifteen exchanges where you and the AI debated three approaches, rejected two, hit an error on the third, pivoted, and finally arrived at something good. That decision‑making process – the most valuable part of the session – just vanishes.
Real problems this creates
- Debugging AI mistakes – When a subtle bug is introduced, you have to hunt through a linear chat scroll to find the exact moment it went off the rails.
- Incomplete code reviews – Your teammate’s PR shows the diff, but you have no idea how the AI got there. Did the developer accept the first suggestion blindly, or iterate carefully?
- Lost knowledge – That clever solution a senior dev built with AI assistance last week is trapped in terminal history, inaccessible to anyone else.
- Fragmented cross‑tool context – Using Claude Code for the backend and Cursor for the frontend splits the context across two separate histories.
Introducing Mantra
I got frustrated enough to build something about it.
Mantra records your AI coding sessions and reconstructs them into navigable timelines.
Think of it like scrubbing through a video, except the video is your coding session.
What Mantra does (and doesn’t do)
- Not a screen recorder – It captures structured data: prompts, responses, tool calls, file changes.
- Timeline navigation – Jump to any point, see what the AI was doing, what files looked like, and understand the flow of decisions.
- Unified view – All sessions from all tools appear in one searchable, structured workspace.
Core idea
AI coding sessions are complex artifacts, and we deserve real tools for reviewing them.
How to use Mantra
The setup is deliberately minimal:
-
Code as you normally do
Use Claude Code, Cursor, Gemini CLI, or Codex. Change nothing about your workflow. Mantra runs in the background and indexes your sessions automatically. -
Open Mantra when you need to look back
You get a unified timeline of all your sessions across all tools. Everything is in one place, searchable and structured. -
Scrub through any session like a video
At each point on the timeline you see:- The prompt you gave
- The AI’s response
- The exact code diff (synchronized)
Jump to the moment a bug was introduced, find that clever pattern from last week, or replay the entire decision‑making process.
No configuration files, no proxying, no monkey‑patching your AI tools.
Use Cases that Stuck
| Use case | Why it matters |
|---|---|
| Debugging AI‑introduced bugs | Identify the exact turn where the AI misunderstood your intent. No more guessing or git bisect through AI‑generated commits. |
| Better code reviews | Reviewers get the narrative: “AI suggested X, developer pushed back, settled on Y because of Z.” |
| Personal recall | “What was that thing I did last Thursday?” – find it in 30 seconds instead of 30 minutes. |
| Cross‑tool search | Claude Code and Cursor sessions appear in the same workspace, searchable together. |
Technical Details
| Attribute | Value |
|---|---|
| Supported tools | Claude Code, Cursor, Gemini CLI, Codex |
| Platforms | macOS, Windows, Linux |
| Current version | v0.9.1 (pre‑1.0, actively developing) |
| Architecture | Local‑first, no cloud required |
| Privacy | Runs locally, no telemetry. Export only when you explicitly choose to. |
| Redaction | Built‑in content redaction for API keys, credentials, proprietary code. |
| Extras | MCP Hub (Model Context Protocol server manager) and Skills Hub (prompt template sharing). |
Get Involved
I am offering lifetime free access to the first 50 users. Not a trial, not a “free tier with limitations” – full access, forever. I want real feedback from developers who push AI coding tools hard, and the best way to get that is to remove the price barrier entirely for early adopters.
- Download Mantra here
- If something is broken, file an issue on GitHub.
- Want to contribute or discuss? Join the community on Discord (link).
Your AI conversations stay on your laptop unless you explicitly export them.
To chat about it, join the Discord.
And if you think the whole premise is wrong, I genuinely want to hear that too.
We are in a moment where AI coding tools are powerful, but the workflows around them are still immature. I built Mantra because I needed it. If you need it too, that is the signal to keep going.