# Quickstart

From zero to a searchable agent history: sync your sessions, find a past prompt, and resume the conversation.

Rendered page: https://agentrelay.com/docs/loop/quickstart
Markdown endpoint: https://agentrelay.com/docs/loop/markdown/quickstart.md

---

This walks you from a fresh install to real value in a few commands: pull your agent history into a local database, search it, drill into a session, and jump back into the exact conversation. Everything here runs locally and needs no account. If you have not installed yet, start with [Install](/docs/loop/install).

## 1. Sync your history

Pull every supported source into the local database. The first run reads everything; later runs are incremental and only read new bytes.

```bash
ai-hist sync
```

This imports Claude Code, Codex CLI, Cursor, OpenCode, Agent Relay (when configured), and compacted trajectory runs. See [Sources](/docs/loop/sources) for exactly what is read and from where.

## 2. Search across everything

Full-text search runs over all sources at once:

```bash
ai-hist search "authentication bug"
```

Results come back with entry IDs in the form `#NNN`. Those IDs are your handle for drilling deeper. Narrow the search with filters when you need to:

```bash
ai-hist search "refactor" --source claude --limit 10
ai-hist search "deploy" --project relay
```

See [Search](/docs/loop/search) for the full filter set.

## 3. Drill into an entry

Take an entry ID from your search results and look at the full prompt, its session, and the exact command to resume it:

```bash
ai-hist show 4521
```

To see what else was happening around that moment — the same session plus nearby entries — use `context`:

```bash
ai-hist context 4521
ai-hist context 4521 --window 15   # widen to a 15-minute window
```

To browse an entire conversation, use `session` with the session ID:

```bash
ai-hist session abc-1234-def
ai-hist session abc-1234-def --full   # no truncation
```

## 4. Resume the conversation

`ai-hist show <id>` prints the exact resume command for the tool that recorded the session. Run it to pick the conversation back up where it left off:

```bash
cd /path/to/project && claude --resume <session_id>          # Claude Code
codex resume <session_id>                                     # Codex
cd /path/to/project && cursor-agent --resume=<session_id>     # Cursor
```

> Copy the resume command straight from `ai-hist show <id>` rather than hand-building it — it already has the right project directory and session ID filled in.

## Next steps

- [Sync](https://agentrelay.com/docs/loop/sync): Keep your history current with continuous watch mode.
  - [Sessions](https://agentrelay.com/docs/loop/sessions): Drill into sessions and resume across every tool.
  - [Stats](https://agentrelay.com/docs/loop/stats): The Burn lens: see where your token spend goes.
  - [Cloud](https://agentrelay.com/docs/loop/cloud): Push for a durable backup and team features when you are ready.