What is Capacitor
Kurrent Capacitor records coding-agent sessions so your team can read the conversation behind a change, continue work in another agent, and review a pull request with its implementation context.
It supports Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI, Kiro, OpenCode, Pi, and Antigravity. Choose which integrations to install and which repositories or directories to record.
How it fits together
Section titled “How it fits together”The CLI installs capture integrations, MCP tools, and skills on your machine. It sends recorded events to your Capacitor server, which stores them and serves the web application at https://<your-workspace>.kcap.ai.
An optional daemon launches hosted agents on its machine and lets you control them through the browser. Your vendor CLI still performs the work on that host.
The CLI is open source. The server is operated by Kurrent for hosted workspaces. System architecture and The installed harness explain the components in more detail.
What gets recorded
Section titled “What gets recorded”Depending on the agent’s available transcript and hooks, a recording includes prompts, replies, tool calls and results, file activity, token usage, subagent relationships, and repository context.
Capture settings limit which work is recorded. Visibility and sharing control who can read it after capture.
Find your way around
Section titled “Find your way around”Choose a repository or project in Scope, then open a product area:
| Area | Purpose |
|---|---|
| Dashboard | See recent work, outcomes, and actions that need attention. |
| Insights | Ask questions about activity, usage, and evaluation trends. |
| Sessions | Read the conversation and evidence behind the work. |
| Agents | Start and continue work on a connected daemon. |
| Work items | Follow an effort across issues, PRs, and sessions. |
| Knowledge | Review docs, approve reusable skills, and maintain shared memory. |
| Evals | Assess recorded work and identify improvements. |
| Flows | Coordinate hosted participants and repeat review rounds. |
Agents can use MCP tools to recall earlier work, save knowledge, attach work items, and keep plans and tasks current. This makes the recorded context available inside the next conversation as well as in the browser.
Follow the Quickstart to record your first session.