Security

Your sessions. Your rules.

Capacitor records what your coding agents do. Here’s exactly where that recording goes.

01

You decide what’s recorded.

  • Never scans your codebase. Capacitor only sees what your agent already read or changed.
  • Opt-in, per coding agent. Nothing is recorded until you switch it on — and you can switch it back off.
  • Exclude any project. Nothing from it is ever sent.
  • Credentials are removed on your machine. Keys, tokens and passwords are replaced before the recording is sent.
  • You can read our code. The part that runs on your machine is public on GitHub.
02

Where it lives.

  • A server of your own. Dedicated to your organization.
  • Hosted in the EU.
  • Encrypted in transit.
  • Never trained on. Never shared.
03

Who can see it.

  • Single sign-on. Only your organization gets in.
  • Visible to your team by default. Make any session private in one click.
  • Personal projects stay private. Work outside your organization’s repositories is only ever visible to you.
  • Search is bounded by your access. Your agent only finds sessions you could already open yourself.
  • Delete any session, any time.
GDPR CompliantEU hostingSingle sign-onEncrypted in transit

What’s in a recording.

When a session is recorded, this is what your server receives — everything you and your agent said, and everything it did:

Every prompt
The full text of every message you send the agent.
Every response
The full text of every agent reply.
Tool calls & results
Commands, file reads, and edits — with timing, output, and the exact changes applied.
Reasoning
The agent’s thinking, alongside the turn that produced it.
Token counts
Input, output, and cache reads and writes — per turn.
Sub-agents
Every agent it spawned, with its own record and token tally.
Repository context
The repository, the branch, and the pull request it relates to.

Every field we capture is named in the docs.

Frequently asked questions.

Does Capacitor upload my source code?
It never scans or uploads your repository — but files your agent read or changed are part of the recording. If a project shouldn’t be recorded at all, exclude it.
What if a credential appears in a session?
It’s removed on your machine first. We match known formats for keys, tokens and passwords, so treat it as a safety net — not a substitute for excluding sensitive projects.
How do I keep a project out?
Exclude it. Nothing from that work is sent.
Who can see my sessions?
By default, teammates with access to the same repository. Work in a personal repo stays private, you can make any individual session private, and you can change your default so every new session starts private.
Where is my data hosted?
The EU. Sign-in runs through WorkOS in the US; the full list of services we rely on is in the privacy policy.
Do you train on my data?
No — and we don’t sell or share it.
Can I delete a session?
Yes. Any session, any time.

Questions about security?

If your team needs detail this page doesn’t cover, ask us directly — we’d rather answer than have you guess. Talk to the team

See it on your own repo.

Start free in minutes. You choose what’s recorded, who sees it, and what gets deleted.

Start free Rather start a conversation? Talk to the team
Built by the team behind KurrentDB — event streams in production are what we do. Coding agents just produce a new kind.
Session recap · yesterday 16:42

Fix flaky auth-token refresh test

Tried mocking the clock first — abandoned after the race persisted. Root cause: token cache TTL rounding. Patched in auth/cache.ts, added regression test. Two approaches rejected, one shipped.

claude-code41 min12.4k tokenscontinue →