FAQ

FAQ

Does tripline send my code or transcripts anywhere? No. It reads local files and makes no network calls. The source is small enough to check yourself.

Does it slow down the agent? Barely. tripline re-reads the session transcript on each tool call. On a 3 MB transcript that takes about 90 ms.

Can it warn instead of blocking? Yes. --warn loop,reread lets those calls run and tells the agent why they look wrong. --ask risky sends destructive commands to your permission prompt instead of blocking them. See Rules. Blocking stays the default because it's the only response every agent honors.

tripline blocked something it shouldn't have. What do I do? Raise that rule's limit or turn it off with a flag (see Rules), and please open an issue with the blocked call.

What if tripline itself breaks? It fails open. Any internal error allows the call.

I'm on a subscription, not the API. Is the cost number meaningful? It's API-equivalent usage, which is useful for comparing sessions and catching runaway ones. It isn't what you're billed.

Which agents are supported? Claude Code and Codex CLI, fully. Cursor in scan and blame; a Cursor hook is planned. opencode is next: it stores sessions in SQLite, which takes more work to read. Each agent's parser lives in internal/transcript, if you want to help.

Is the risky rule enough to keep me safe? No. It catches common destructive commands, but it's a pattern list, not a sandbox. Keep permission prompts on for commands that matter.