Quickstart

What is Capacitor

Capacitor records every coding-agent session your team runs — Claude Code, Codex, Gemini CLI, Pi, OpenCode, Antigravity, GitHub Copilot CLI, Cursor, and Kiro — and surfaces them in a real-time dashboard. A small CLI installs hooks into your agents that stream each session to your Capacitor server as it runs. From there you can replay a session, recap it, hand it to another agent, share it, and use it to review the pull requests it produced — with full transcripts queryable from the CLI, the dashboard, and from inside the agent itself.

Capacitor records your agent sessions

Once its hooks are installed, Capacitor captures your coding-agent sessions automatically — full conversation transcripts, tool calls, and diffs stream to your Capacitor server as you work. You can configure it to only capture certain sessions: set a default visibility, mark individual sessions private, or exclude repos from capture entirely — see Visibility & sharing. For how capture works under the hood, see System architecture.

Setup Capacitor

1. Install the CLI and create your workspace

1. Install the CLI and run the setup wizard:

   npm install -g @kurrent/kcap
   kcap setup

2. Sign in when your browser opens Kurrent’s hosted sign-in. First time here? Choose Sign up at the bottom of the form, then return to the terminal.

3. If no workspace is linked to your account yet, the wizard offers to create one — “Create one now?” Name your organization and pick your workspace URL:

   https://<your-account>.kcap.ai

   Provisioning takes a few minutes; the wizard waits for the workspace to come online, then carries on. Workspaces created this way start on the Free plan. (Already have a workspace? The wizard discovers it and skips straight ahead.)

4. Let the wizard finish: it asks for your default session visibility, then installs hooks for every supported agent it detects (Claude Code, Codex, Gemini CLI, Pi, OpenCode, Antigravity, Copilot CLI, Cursor, Kiro).

5. Confirm it worked:

   kcap whoami
   kcap status

   You should see your signed-in identity and a reachable server.

6. If you use Codex, start the Codex CLI and accept the new Capacitor hooks.

Note

For the full wizard walkthrough — including slug rules and every hooks option — see Setup CLI. For what each visibility choice means, see Visibility & sharing.

Caution

If you use Codex in a later step, Codex won’t fire its Capacitor hooks until you accept the trust prompt.

2. Sign in to Capacitor

Open your new workspace in the browser to confirm it’s live.

1. Open your workspace URL:

   https://<your-account>.kcap.ai

2. Sign in through Kurrent’s single sign-on (SSO) — the same account you just used in the wizard. Anyone who’s a member of your Capacitor organization can sign in — there’s nothing to install and no separate access request.

3. Import your history to Capacitor

Import your past coding-agent sessions into Capacitor.

1. Start the import:

   kcap import

2. At What would you like to import?, choose Specific repository, then pick a repo you’ve worked in from the list.

3. Read the summary it prints and confirm.

Note

Only repos with captured sessions appear in the list, so your pick always has something to import. You can run the command again at any time — it only uploads what’s missing.

4. Open Capacitor Server and see your sessions

See past imported sessions in the dashboard.

1. Open your Capacitor server in a browser:

   https://<your-account>.kcap.ai

2. Select the Sessions tab and you should see all the sessions that you have imported.

Note

Capacitor displays your session chat history, events, traces, token costs, and other session details.

Multi-agent Hand-off

1. Open a session and copy its ID

Pick a past session in the dashboard and copy its ID.

1. Open your Capacitor server in a browser:

   https://<your-account>.kcap.ai

2. Select the Sessions tab:

3. Open any session, and note which agent it ran with — you’ll hand it to a different agent next.

4. Click the copy icon beside the session title.

2. Hand the session to another coding agent

Hand the copied session off to a different agent.

Note

This step moves a session between agents, so you’ll need both Claude Code and Codex installed.

1. In the session’s repository, start the other agent — Codex if the session ran in Claude Code, Claude Code if it ran in Codex:

   cd <the-repo>
   codex            # or: claude

2. Prompt it to pick up where the session left off, pasting the ID you copied:

   Recap session <paste-the-id> and continue where it left off.

3. Continue working as if it were the same agent — it picks up with the prior session’s decisions and ruled-out approaches already in context.

Share the session with a teammate

Give a teammate access to the session from the dashboard.

1. Open the session in the dashboard and click the link 🔗 icon next to the session name.

2. Under Teams & people, choose a teammate or a team from your organization.

Note

They can now open the session — and, as you’ll see in the next step, query its transcript from their own machine.

Review a pull request

Review a pull request using the sessions that produced it.

1. In the dashboard, find a session that’s linked to a pull request — the PR shows in the session’s metadata. Copy that PR’s URL from GitHub.

Note

No PR-linked sessions among the ones you imported? This is the step to come back to once you’ve run a session whose branch becomes a pull request.

2. In the CLI, run:

   kcap review <pr-url>

3. Ask questions about the change to understand the reasoning behind it — the answers come from the session transcripts, not just the diff.

What you’ve done

You’ve taken one session through the entire Capacitor loop: captured, imported, handed off, shared, and used to review real work.

Each step has a page that goes deeper:

• Setup Server — the one-time admin setup.
• Setup CLI — the full setup wizard.
• Import your history — other import scopes.
• Session recap — recaps in the terminal and the in-agent skill.
• Visibility & sharing — every way to control who sees a session.
• PR review — the MCP tools behind the review.

And to go beyond the tutorial:

• The dashboard — what each tab shows.
• Session recall — search every session from inside your agent, not just one pull request’s.
• CLI commands — the full command surface.