paper

Instead of treating each agent run as a temporary terminal interaction, paper records the session as structured data: prompts, turns, tool calls, token usage, cost, duration, status, and metadata.

paperctl CLI captures the work as it happens locally. paper console turns that captured work into durable session history your team can view, search, inspect, and learn from over time.

How paper creates value

Capture your work        → run agents through the paperctl CLI
Understand what happened → inspect sessions in the console; search them from the CLI
Reuse what worked        → turn the best sessions into versioned Skills

Capture is the entry point: run an agent through paperctl start (or shell routing) and the session — prompts, turns, tool calls, usage, cost, status — lands in paper console as structured history. How capture works traces the whole path. Understanding starts with one session (“how did this happen?”) and grows into Insights across many. Reuse is where the history pays compound interest: the sessions worth repeating become Skills your whole team applies.

Start here

Which docs should I use?

• Quickstart — capture your first session and prove the value, in about ten minutes.
• CLI docs — install, authenticate, capture with Claude Code, Codex, or Pi, and work with sessions from the terminal, including search.
• Console docs — read session details, navigate the team workspace, and interpret Insights.
• Skills docs — publish, install, and roll out reusable knowledge from captured work.

Going further

Once you’re capturing sessions, these are the next steps from everyday use to getting the most out of paper:

• Custom backends — customize routing, set model allowlists, or share one team-managed key.
• Publish your first skill — turn a captured session into a reusable skill your team can apply again.
• Team workspace — switch orgs and gateways, filter the team’s sessions, and read the aggregate dashboard.