Skip to main content
By the end of this guide your agents run on a plan you already pay for, such as Claude Pro or Max, a ChatGPT plan or GitHub Copilot, instead of on an API key or Reliant credits. Reliant’s servers make the model calls under your account with the provider. Signing in only has to hand Reliant a login, and that hand-off is the one part that needs your own computer.

Before you begin

  • A paid plan with the provider.
  • The Reliant desktop app, or the web app with the reliant CLI installed on the computer you browse from (Installation).
1

Choose the provider

Open Settings → AI providers, click Add provider, and pick one from the subscription group:Each plan unlocks only its own models. The Providers column of the Models Reference lists them as claude, codex, antigravity and copilot.
2

Sign in

Claude, ChatGPT and Google send your browser back to localhost when you approve, so something on the computer you are browsing from has to receive it.
  • Desktop app: nothing to do. Its machine receives it.
  • Web app, with reliant daemon start running on this computer: the app asks that machine to receive it.
  • Web app, otherwise: the sign-in panel asks for the helper. Run it and leave it open until you are connected:
terminal
terminal
The panel notices it within a few seconds and shows the sign-in button. The helper only passes the authorization code to the app; your login is exchanged and stored by Reliant, with your account. Check it is up with curl http://127.0.0.1:19284/health:
terminal
GitHub Copilot needs none of this: Reliant shows a code, you enter it on GitHub’s site and approve.
3

Check the account is connected

The provider’s card in Settings → AI providers shows the account. From a terminal, reliant providers list shows every account your chats can run on, and who pays for each:
terminal
terminal
You can connect more than one account per provider (Add another Claude account); a chat uses one at a time, and when that one hits its usage limit the chat moves to another or waits for the reset.
4

Use it

The models it unlocks now appear in a chat’s model picker. A workflow that asks for a tier, such as tags: [flagship], gets the best model in that tier from the providers you have connected, so built-in workflows use the subscription with no change (Model configuration).

What this does not do

  • It doesn’t lift your plan’s limits. Usage counts against the subscription, under your account with the provider, and its terms and limits apply. At a limit the run doesn’t fail: the chat shows which provider is limited and when it resets, and carries on at the reset. Retry now and Stop are there, and another connected provider can take over in the meantime (AI providers).
  • It doesn’t run model calls on your computer. Reliant’s servers call the provider; your computer is only needed to sign in.
  • In an organization, it isn’t spent without you. A chat in an organization asks before it moves from the organization’s AI accounts to your own; reliant workflow funding <execution-id> shows the question, and only a person in the app can answer it. An automation turned on from the CLI spends only the organization’s accounts.
  • The CLI can’t connect a provider. Accounts are added in the app; reliant providers only lists them.

Troubleshooting

Run reliant auth serve on the computer your browser is on, and keep it open until the panel shows the sign-in button.
Your machine is a cloud machine or another computer, so the localhost redirect can’t reach it. Run reliant auth serve here.
A helper, or a daemon that opened one, is already running on this computer, so you don’t need another. If another program holds port 19284, stop it: the web app looks for the helper on that port only.
Reliant is signed out of that account, so chats can’t use it. Click Reconnect on the account in Settings → AI providers and sign in again.

Next steps

Run a local LLM coding agent

Add Ollama or vLLM beside your subscription, one step each.

Use a different model for each step

Pin a step to a model your plan unlocks.

Write your first multi-agent workflow

Put the subscription to work on a planner, an implementer and a reviewer.

AI providers

API keys, Reliant AI and every source side by side.