Keep your agent history useful.
Hansard turns conversations from coding agents into a durable, searchable archive that stays under your control.
Install with your agent
Section titled “Install with your agent”Copy this prompt into a coding agent with access to the machine's terminal:
Install and fully set up Hansard on this machine. Check for Node.js 20 or newer,then install the published CLI with npm install -g hansard. Read the installedhansard/skills/setup/SKILL.md (find the package directory with npm root -g).Run the bounded setup preview and check disk space first. Use local storage,import the last 30 days from detected local sources, install recall for supportedagents already present, enable same-project agent collaboration, and start thewatcher with login startup. Open the viewer and verify setup, imports, recall,and watcher health. Preserve any existing archive and preferences. Leave remotesync, sharing, telemetry, CLI usage tracking, and agent access to memories offunless already configured. Ask only for missing choices or a blocking issue;complete the authorized steps and report anything still pending.The setup guide ships with the npm package. This prompt authorizes local setup, a 30-day import, recall integration, and background archiving. Change those choices in the prompt before sending it when needed.
One archive, across agents
Section titled “One archive, across agents”Hansard imports conversations from supported coding agents, preserves the source material needed for future reprocessing, and builds provider-independent records for browsing and recall.
Install in minutesInitialize a local archive, verify the setup, and open the viewer.
Understand the archiveSee what Hansard stores, what remains local, and how the pieces fit together.
Built for durable recall
Section titled “Built for durable recall”- Local-first by default. Normalized archives, search indexes, memories, and source backups live on the machine unless sync or sharing is configured.
- Provider-independent. Search Codex, Claude, Gemini, Cursor, and other supported sources through one archive.
- Reprocessable. Raw source preservation and versioned normalized records allow improved parsers and views without discarding history.
- Agent-ready. Recall integrations let supported coding agents search and continue from past work without manually locating transcript files.