> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usesyndicate.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Providers and models

> Supported providers, available models, model selection, and auth behavior.

Syndicate does not ship its own models. Agents run on the provider accounts you connect: Claude, OpenAI (Codex), Google Gemini, and GitHub Copilot. Connect one provider or all of them; agents in the same team can use different providers.

<Frame caption="Provider cards show connection status, available models, and account actions">
  <img src="https://mintcdn.com/syndicate-e378bd26/tUHYLKtcaVnFhlNX/assets/screenshots/providers/providers-connected.png?fit=max&auto=format&n=tUHYLKtcaVnFhlNX&q=85&s=d30f22de0d0ef3f1780c253f441c3d99" alt="The Providers settings screen showing connected provider cards and available models" width="1920" height="1128" data-path="assets/screenshots/providers/providers-connected.png" />
</Frame>

## Supported Providers

| Provider           | Account used                                              | Auth flow                                       |
| ------------------ | --------------------------------------------------------- | ----------------------------------------------- |
| **Claude**         | Your Claude account                                       | Browser sign-in, paste a code back into the app |
| **OpenAI (Codex)** | Your ChatGPT account with Codex access                    | CLI sign-in                                     |
| **Google Gemini**  | Your Google account                                       | CLI sign-in                                     |
| **GitHub Copilot** | Your GitHub account with a Copilot plan (free tier works) | CLI sign-in                                     |

Each provider works through its CLI. If the CLI is missing, Syndicate offers **Install CLI** and installs it with its bundled runtime.

How-to setup: [Connect providers](/guides/connect-providers).

## Available Models

Models appear only for providers that are connected. Current lineup:

| Provider       | Models                                                                                                                                  |
| -------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
| Claude         | Fable 5, Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 5, Sonnet 4.6, Haiku 4.5                                                                  |
| OpenAI (Codex) | GPT-5.5, GPT-5.4, GPT-5.4 Mini                                                                                                          |
| Google Gemini  | Gemini 3.5 Flash, Gemini 3.1 Flash Lite, Gemini 3.1 Pro (preview), Gemini 2.5 Pro, Gemini 2.5 Flash                                     |
| GitHub Copilot | Auto (via Copilot), Claude Sonnet 4.5 (via Copilot), Claude Sonnet 4 (via Copilot), Claude Haiku 4.5 (via Copilot), GPT-5 (via Copilot) |

Model lists update as providers ship new versions. The in-app selector reflects what your connected CLIs support.

## Model Selection

There are two levels:

1. **Default model**: Settings, Providers, **Default Model For New Agents**. Used for new teams and solo agents.
2. **Per-agent model**: set during team creation or from the agent configuration panel.

If a provider disconnects, agents pointing at its models resolve to the first available model from your remaining providers.

### Choosing Models For A Team

* **Atlas**: your most capable model. Planning and coordination benefit most from reasoning quality.
* **Coders**: a strong code model.
* **Reviewers, writers, support roles**: a balanced model.
* **High-volume monitors**: a faster model.

The built-in [team templates](/reference/team-templates) follow this pattern.

## Provider Readiness

A provider is usable when:

1. The CLI is installed.
2. The CLI is authenticated.

The Settings card reflects this: **Install CLI**, **Connect**, or **Connected** with model list and account actions. **Check** / **Refresh** force a re-probe.

## Credentials

Each provider's CLI manages its own credentials:

| Provider       | Credentials                                 |
| -------------- | ------------------------------------------- |
| Claude         | `~/.claude/.credentials.json`               |
| OpenAI (Codex) | `~/.codex/auth.json`                        |
| Google Gemini  | `~/.gemini/oauth_creds.json`                |
| GitHub Copilot | Managed by the Copilot CLI's GitHub sign-in |

Syndicate watches these locations during sign-in, which is why terminal-based logins complete automatically. Deleting a credential file signs that provider out.

## Provider Behavior

### Claude

* Primary flow is browser sign-in with a paste-back code.
* If the sign-in endpoint is rate-limited, Syndicate falls back to terminal login.
* Codes expire after about 15 minutes.

### OpenAI (Codex)

* Requires a ChatGPT plan with Codex access.
* If access is missing, use Claude/Gemini or upgrade your ChatGPT plan.
* Reconnect from Settings to switch accounts.

### Google Gemini

* Sign-in runs through a terminal flow, including two-factor prompts.
* Revoking CLI access in your Google account signs it out.

### GitHub Copilot

* Sign-in runs through a terminal flow (`copilot login`) with your GitHub account.
* Requires an active Copilot plan; the free tier works. The **Auto (via Copilot)** model lets Copilot pick a model for you, which is the most free-tier-friendly choice.
* Copilot's models are hosted Claude and GPT backends, labeled "(via Copilot)" in selectors so they are not confused with the direct Claude and OpenAI providers.

## Expiry During Work

If credentials expire mid-session, agents using that provider stop and an amber banner offers **Reconnect \[Provider]**. Reconnect and re-prompt affected agents; queued work is preserved.

## Account Auth

Your Syndicate account session is stored encrypted on-device and restored on launch. Sign-out clears the local encrypted session first, then revokes the token best-effort. Account tiers may limit concurrent devices.