FAQ

Short answers to the questions that keep coming up. If yours is not here, open an issue on the repo — we read every one.

Is SenateLab open source?

The app is MIT-licensed and the source lives at github.com/KnotekBerzas/senatelab. Build locally, audit the IPC surface, ship a fork — all fair game.

Does SenateLab send my code to the cloud?

Only what you ask the agent to do. The app itself runs locally; the data store is on your disk; no telemetry is collected by default. When you send a message, the Codex CLI makes a request to whichever provider you configured — that call follows the provider's own privacy policy. Using a local provider (Ollama, self-hosted) keeps everything on-device.

Can I self-host?

There is no server to host — SenateLab is a desktop app. If by "self-host" you mean "run without any external provider", yes: pair it with Ollama or any OpenAI-compatible endpoint running on your own infra.

Which models are supported?

Any model your configured provider exposes. OpenAI-compatible endpoints pass the model name through as a string; Ollama uses whatever you have pulled. We test against a small matrix on each release — the full list ships in the changelog.

How do I add a custom provider?

Open Settings → Providers → Add custom. Enter the base URL and an API key if required. The endpoint must speak the OpenAI Chat Completions protocol (streaming SSE, the usual JSON shape). A Test button makes one roundtrip before you commit.

How is this different from running Codex CLI in separate terminals?

Three things: coordination (agents talk to each other and to you through the canvas), isolation (scoped folders and explicit guardrails rather than "please don't do X"), and memory (persistent reports and summaries per agent). Running four terminals gets you four processes; running four agents in SenateLab gets you a team.

What happens if an agent goes rogue?

It cannot — within the sandbox it was given. Folder access, guardrails, and approval prompts stop unintended actions before they reach the filesystem or the network. If an agent produces bad output, you see the status go red, you see the log, and you can cancel with Esc. Memory can be wiped per agent; the agent itself can be deleted with a click.

How do I report a bug or request a feature?

Open an issue: github.com/KnotekBerzas/senatelab/issues. Bugs get a label and a triage comment within a business day. Feature requests get a discussion thread — even the ones we do not build right away are useful signal.