Every conversation you have with an AI coding CLI lands in a dot directory somewhere. Command Center indexes all of them into a single searchable timeline: skim what you worked on each day, deep search full transcripts, replay any session, and copy a ready to run resume command.
Tools you don't use are skipped automatically. Every location can be overridden with an env var.
CLAUDE_HOME · OPENCODE_HOME · KIMI_HOME · COMMANDCODE_HOME · T3CODE_HOME · ANTIGRAVITY_HOME · GROK_HOME
Every session from every tool in one reverse chronological list, grouped by day, filterable by tool and project.
SQLite FTS5 full text search across conversation content, with snippets and highlighting. The index updates incrementally; only changed sessions are re-read.
Replay any conversation: user messages, assistant responses, collapsible tool call groups, and turn summaries.
One click copies the exact CLI command to resume a session in its original tool and project directory.
A GitHub style contribution graph of sessions per day, with per day token usage on hover. Input, output, and cache totals per session.
Search and inspect your history from the terminal, including flow: reconstruct how a project was built from its transcripts.
No database server, no build step. The fastest path is Cove, a zero config local dev environment for macOS. Any server that routes through index.php works too.
Everything the dashboard does is available as JSON under /api. Script it, pipe it, build on it.
Command Center only reads local transcript files and builds its search index in a local SQLite database. Nothing is uploaded anywhere.
No analytics, no phone home, no external requests from the backend. The transcript data and FTS index live in data/, which is gitignored.
Your conversation history is sensitive. Run it locally only; the app ships with no auth, so don't put it on a public server without adding some.