FredrinMost tools in this category resell you tokens. Fredrin does not. A flat-rate coding subscription is dramatically cheaper than the same work metered through an API, and that saving should be yours.
Two ways to power a Worker
Settings → Agents offers a choice per workspace.
Bring your own
The default, and the one we recommend. Fredrin injects no credential at all. It spawns the coding CLI you already have installed, and that CLI uses its own on-machine login.
If you can run claude in your terminal and it works, Fredrin can run Workers. There is nothing to paste and nothing to configure.
This covers the subscription you are already paying for: Claude Pro or Max, ChatGPT Plus or Pro for Codex, Cursor, GitHub Copilot.
Managed
Fredrin provisions and injects a spend-capped credential of its own, billed per token against your workspace's monthly included allowance, with postpaid overage beyond it.
Use this when you have no coding subscription, or when you want a workspace to run without every member holding their own.
Exactly one of the two is active per workspace, and switching is a single click.
Where credentials live
Settings → Agents has no field for a provider token, and never reads one. That is deliberate. Under bring-your-own, the credential is the one your CLI already holds, and there is nothing for Fredrin to store.
Credentials that Fredrin does hold, for organisations whose policy forbids subscription login, live in Settings → You → Connections: one surface for the question "what am I signed in to", rather than two entry points answering it differently. A stored credential is encrypted at rest and injected into the Worker's spawn environment, so the CLI authenticates by token instead of by login, and it takes precedence over the CLI's own session.
There is deliberately no "sign in to your provider through Fredrin" option. CLI login already is OAuth underneath, and subscription OAuth is not exposed to third parties. Anyone offering you that flow is doing something you should look at closely.
What Fredrin charges for
The app. That is the whole model.
- No per-token bill.
- No markup on model usage.
- No seat charge for agents, only for people.
The exception is Managed mode above, where Fredrin is buying the tokens and passes through metered usage against your allowance.
See pricing for the plans, and workspaces and roles for which entitlements come with which tier.
Choosing a model
Model and effort are resolved per ticket, so a cheap mechanical change and a hard architectural one do not have to run on the same setting. The defaults are sensible; the picker is there when you want it.
Choosing your agent covers which CLIs are selectable today and what each one supports.
Next steps
- Choosing your agent - runtimes, models, and effort levels.
- Workspaces, members and roles - who pays for what.
- Privacy - what leaves your machine, and what never does.