Use cases

Each section answers one question developers ask about their coding agents, with the command or app view to use and a link to the full guide. Hansard runs on your computer and needs no account. Its archive lives in ~/.hansard.

How do I search my past Claude Code, Codex, and Cursor conversations?

Hansard imports the history of every coding agent on your computer into one archive and searches all of it at once. In the app, open Sessions and press /. From a terminal, run hansard history "rate limiter", and narrow the results with --repo, --provider, --tool, --command, --file, or --commit. See Search.

Which agent conversation wrote this line of code?

Run hansard why <file>:<line>, such as hansard why src/server/auth.ts:120. Hansard matches the line against the file edits in your archived conversations and names the session that last wrote it, with the prompt that led to the edit. In a Git checkout it also shows the commit that last changed the line. See Search.

How do I manage memory files like CLAUDE.md and AGENTS.md across agents?

Hansard backs up the memory and instruction files your agents keep, such as CLAUDE.md, AGENTS.md, GEMINI.md, Codex memories, and Cursor memories and user rules. While the watcher runs, it checks them every 15 minutes and keeps a revision history of each file. Search every agent’s memories in Memories or with hansard memory search, edit a memory file in the app, and run /unify-memories in Claude Code to merge duplicates and resolve contradictions across agents. See Memories.

How do I give my coding agent memory of earlier conversations?

Install recall, which connects the agent to Hansard’s MCP server. Run hansard integrations add-to claude, or name another agent such as codex or cursor. The agent can then search past conversations with search_past_sessions, read one with get_past_session, and search memories from every harness with search_memories. In Claude Code, type /recall and a topic. See Recall in your agents.

How do I keep Claude Code from deleting my conversation history?

Claude Code deletes CLI transcripts older than 30 days by default, controlled by cleanupPeriodDays. Hansard’s watcher imports each new session as you work and keeps a copy of the original file, so the conversation stays in your archive after Claude Code removes it. Run hansard init to import your existing history and set up the watcher. See Background archiving and Harness settings.

How do I back up or export my coding agent chat history?

The archive is a backup: it keeps each conversation as a readable transcript plus a copy of every original file the harness wrote. To export one conversation, run hansard export <session-id> --out <folder>. To keep a copy of the whole archive on storage you own, use Personal Sync. See Browse conversations.

How do I import my ChatGPT, Claude.ai, and Gemini conversations?

Request a data export from the service, then import the file it sends you. In the app, open Settings → Imports and select Import export…. From a terminal, run hansard import chatgpt <export.zip> or hansard import claude-web <export-folder>. See ChatGPT, Claude.ai, and Gemini exports.

How do I track token usage and cost across Claude Code, Codex, and Cursor?

Stats totals conversations, messages, tokens, and estimated spend at API prices across every harness and model, for any time range. Open Stats in the app, or run hansard stats. To follow your subscription’s plan limits instead, use Accounts and plan usage. See Stats.

How do I continue a Claude Code conversation in Codex or another agent?

Open the conversation in Sessions, choose Continue in…, and pick an installed harness. Hansard starts a new conversation there with a summary of the original. Inside an agent, /continue-from loads an earlier conversation into the one you have open. See Continue in another harness.

Can coding agents in the same project message each other?

Yes. Live Codex and Claude Code sessions with Hansard’s tools, working in the same project, can find each other and exchange notes and questions. Messaging is on for the same project by default, and Boards in the app shows the threads. An agent can also start helper agents in another harness, such as a Codex session that asks Claude Code for a second opinion. See Agent messaging and Helper agents.

How do I give a project a long-running coordinator agent?

Open the project’s overview in the app and start its coordinator: a persistent Codex or Claude Code conversation that keeps the project’s goals and hands work to helper agents. It receives updates about recent conversations and blockers across the project. See Projects and coordinators.

How do I answer broad questions about past work with agents?

Deep Recall plans several searches, reads the strongest conversations in full, and writes an answer that cites each one. Use it for questions such as which decisions were reversed or how an approach changed. Open Deep Recall in the app, run hansard deep-recall, or type /deep-recall in Claude Code. See Deep Recall.

How do I sync coding agent history between my computers?

Personal Sync backs up your whole archive to a bucket or shared folder you own and brings each device’s conversations into every installation. Open Settings → Personal Sync, or run hansard personal-sync. See Personal Sync.

How do I share agent conversations with my team?

A Team shares whole conversations from chosen repositories through a bucket or shared folder your team controls. Teammates receive them read-only, labeled with the person and device. To post recaps and daily digests to a channel instead, use Slack. See Teams.

Which skills, plugins, and MCP servers does each agent have installed?

The Skills & plugins view lists the skills, plugins, MCP servers, and connectors every harness on your computer has configured, with earlier versions of each skill. Run hansard skills for the same inventory in a terminal. See Skills, plugins, and MCP servers.

How do I reduce Claude Code permission prompts?

Run hansard permissions suggest in your project folder. It replays the shell commands your agents ran against your Claude Code Bash(...) allow rules and suggests rules for the frequent commands they miss. See Permission suggestions.

How do I track plan limits across several Claude and Codex accounts?

The Accounts view keeps your agent and browser accounts in one place, shows how much of each plan’s limits you have used, and can choose which account a new session uses. Run hansard account for the same from a terminal. See Accounts and plan usage.

How do I see how my projects and conversations connect?

Graph draws your archive as a network of projects and conversations, linked by shared files, commits, continuations, and agent messages. Graph is available only in the app. See Graph.

Which coding agents does Hansard support?

Hansard imports Claude Code, Codex, Cursor, GitHub Copilot, Antigravity, Aider, Cline, Roo Code, Kilo Code, OpenCode, Goose, Amp, Zed, Warp, Kiro, Devin, Factory Droid, Jules, Grok Build, and more, plus ChatGPT, Claude.ai, and Gemini exports. It also keeps history from retired tools such as Gemini CLI and Windsurf. Each harness has a guide with its source IDs and what it records. See Supported harnesses.

Does Hansard send my conversations anywhere?

Importing, search, Stats, memories, and the app run on your computer without a Hansard account. Work classification, which is on by default, sends a digest of each finished conversation to the Claude or Codex CLI you are signed in to. Personal Sync, Teams, Slack, and Deep Recall send data only after you set them up or ask. See What leaves your machine.

Is Hansard free?

Yes. The CLI installs from npm with npm install -g hansard under the MIT license, and the desktop app for macOS and Windows is a free download. Neither needs a Hansard account. See Install the CLI or Install the app.