Introduction

Hansard is a local-first archive, search, and recall tool for conversations with AI coding agents. It brings histories from supported providers into one durable format while retaining the source material needed for later reprocessing.

What Hansard provides

• A normalized archive of conversations, events, metadata, and attachments.
• Terminal search with repository, provider, tool, command, and time filters.
• A local web viewer and a separately installed native macOS app.
• Recall and continue-from integrations for supported coding agents.
• Activity, model, tool, token, and spend summaries when source data permits.
• Optional personal sync, signed Team sharing, and Slack notifications.

How the pieces fit together

1. Importers read supported local stores or explicit web exports.
2. The archive preserves raw sources and writes normalized session files.
3. The index makes conversations and archived memories searchable.
4. The watcher imports newly written conversations in the background.
5. Viewer and recall surfaces read the same normalized archive.

Data boundary

Hansard does not require a hosted account for local import, search, or viewing. Source history and normalized archives stay local unless a feature that moves data—such as remote sync, Teams, Slack notifications, or an explicitly hosted viewer—is configured.

Redaction can protect normalized transcripts before import, but byte-preserved raw source copies remain local and unredacted. Treat archive backups with the same care as the original provider data.

CLI and native app

The npm package supplies the hansard CLI and browser viewer. On macOS, hansard app can activate the separately installed native application. If the native app is unavailable, the command falls back to the local browser viewer.

Continue with the quickstart to create the first archive.