Quickstart

Run Chronicle and reach your first time-travel moment — clicking a message and watching your code snap back to how it looked at that instant — in a couple of minutes.

Chronicle's core trick is simple to experience but hard to forget: it lines up every message in an AI coding session against your Git history, so any point in the conversation becomes a window into the exact state of your code. Here's the whole loop in a few seconds:

Chronicle — Playback● LIVE

User

Assistant

Tool Call

User

Assistant

server/git.js

1export function commitAt(dir, ts) {

2 // TODO: pick a commit

2 const hash = git(dir, ['rev-list', '-1']);

3 return describeCommit(dir, hash);

4}

5 

6function treeAt(dir, commit) {

Click any message in your sessionChronicle rewinds your code to that exact momentToggle the diff to see what changedScrub the timeline like a video

No account, no API key, no network — Chronicle reads the logs your AI tools already wrote to disk, entirely on your machine.

1 · Launch it

npx chronicle-cli

That's the whole install. npx fetches Chronicle, starts a local server (default http://localhost:41730), and opens your browser to the dashboard automatically. Requires Node.js 24+ — see Installation if you need to check or upgrade your Node version, or want the --port/--no-open flags. Press Ctrl-C in the terminal to stop it whenever you're done.

2 · Land on Home

The dashboard opens on Home: a day-grouped stream of your most recent sessions across every tool, with search up top and a project rail on the right. First run, this is empty — the next step fills it in.

3 · Import your sessions

1. Click + Import Sessions (top bar, available from any view). Chronicle scans the standard log locations for all four supported tools — Claude Code, Codex, Cursor, and OpenCode — and shows what it found on your machine, no configuration required.
2. Pick a source — Claude Code is usually the richest if you have it installed.
3. The wizard lists sessions with NEW / Partial / Imported badges (new ones are pre-selected). Hit Start Import — it's read-only, so your original logs are never touched.

After the first import, Chronicle keeps itself in sync automatically — new sessions and updates to in-progress ones are picked up in the background as you use your AI tools, with no manual re-import needed. See Supported tools for the full tool matrix and log locations.

4 · Pick a Git-backed project

Chronicle time-travels through Git commits, so open a session whose project is a Git repository with some history. The more commits it has, the more precisely Chronicle can reconstruct code between messages. A project with no repo still plays back the conversation — you just won't get the code-snapshot pane.

5 · Time-travel

1. From Home, click a project card on the right rail, then click any session — or open one straight from the recent-sessions stream.
2. The session opens on Overview — stats, cost & usage, agent-active time, and a Subagents card if the session used any. Switch to Playback from the left rail (or press ⌘2).
3. Click any message. The middle pane rebuilds your file tree and file contents as they were at that moment, resolved to the nearest preceding commit. Changed files are green-dotted and auto-selected. Press D for the diff.
4. Drag along the timeline (bottom) to scrub the whole session and watch the code evolve commit by commit.

That's the "aha" — the whole point of Chronicle.

6 · Explore Insights

Click Insights in the sidebar for the cross-project view — three tabs: Overview (spend, tokens, agent-active time, tool mix, and a top-sessions-by-cost table across everything you've imported), Explore (a pivot table — slice spend/tokens/requests by model, project, source, tool, skill, or subagent), and Content (what's actually filling your context — tool results, skills, and subagents, as a share of total tokens). The same three tabs exist scoped to a single project (its Overview / Explore / Content / Sessions tabs) and scoped to a single session (from that session's Overview, drill into Content).

Local-first: Every step ran entirely on your machine. Chronicle made no model call and no network request — it parsed local logs into a local SQLite database and reconstructed code from your own commit history. Nothing about viewing a session leaves your laptop. The one feature that does run a model is Ask, which is off by default; turned on, it runs claude -p locally on your own Claude subscription, once per question you type.

Related

• Installation — the Node requirement, CLI flags, and where data lives.
• Supported tools — the tool matrix, log locations, and configuration.
• How it works — the ingestion pipeline, the Git snapshot engine, and the Insights engine in depth.