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

# Model Provider Quickstart

# Model Provider Quickstart

Fased uses one provider registry for onboarding, CLI, Agent setup, Chat, tasks,
and channel-routed Agents. The normal browser surface is **Agent > Models** for
the selected Agent.

## Quick Start Path

1. Open **Agents**, select an Agent, and use **Agent > Models** to add a
   credential, sign in, paste a token, or configure a local/manual endpoint.
2. Choose that Agent's primary/fallback/task model roles on the same page.
3. Use **Chat** to test the Agent. Chat can override the model for the current
   session when the route is usable.
4. Use **Agent > Channels** to route external apps to that Agent.
5. Use **Usage** when you need local token totals grouped by provider, model,
   Agent, session, task, or channel.

```mermaid theme={"theme":{"light":"min-light","dark":"min-dark"}}
flowchart TD
  start["Open Agent"] --> auth["Add provider auth"]
  auth --> roles["Set model roles"]
  roles --> chat["Test in Chat"]
  chat --> channels["Route channels"]
  chat --> usage["Review Usage"]
  channels --> usage

  classDef entry fill:#10151f,stroke:#38bdf8,color:#e0f2fe;
  classDef setup fill:#15110a,stroke:#f59e0b,color:#fff7ed;
  classDef run fill:#102016,stroke:#22c55e,color:#ecfdf5;
  class start entry;
  class auth,roles setup;
  class chat,channels,usage run;
```

## Pick The Provider Shape

<CardGroup cols={2}>
  <Card title="Hosted API or sign-in">
    Start with [OpenAI](/providers/openai), [Anthropic](/providers/anthropic),
    [Chutes](/providers/chutes), or another hosted provider page.
  </Card>

  <Card title="Router or aggregator">
    Start with [OpenRouter](/providers/openrouter),
    [Vercel AI](/providers/vercel-ai-gateway), [LiteLLM](/providers/litellm),
    or [Cloudflare AI Gateway](/providers/cloudflare-ai-gateway).
  </Card>

  <Card title="Local model">
    Start with [Ollama](/providers/ollama), [LM Studio](/providers/lmstudio),
    or [vLLM-compatible](/providers/vllm).
  </Card>

  <Card title="Custom endpoint">
    Use [Custom Provider](/providers/custom) when the built-in provider shapes
    do not fit your endpoint.
  </Card>
</CardGroup>

## Catalog Maintenance

The provider's authenticated model catalog is the availability authority.
Fased's checked-in catalog supplies reviewed capabilities, recommendations, and
role rankings; it is not a substitute for account access. **Providers**,
**Agent > Models**, Chat, and Tasks use the same Gateway response, so their
available counts and selectable model IDs should match for the selected Agent.

Ollama, LM Studio, vLLM-compatible servers, and LiteLLM are discovered from the
user's configured endpoint. They have no hard-coded placeholder model.

Use `fased providers refresh` to compare the checked-in provider registry with
live or reviewed source catalogs. Use `fased providers refresh --write-review`
before applying model or capability changes.

The repository also runs a nightly credentialed drift review. Drift creates a
pull request for human review; it never changes a release or recommendation
silently.

Use `fased providers models add/remove` for local or custom model entries on the
current machine.

See [Model providers](/concepts/model-providers) for the exact refresh and
manual-model commands.

Provider docs intentionally list only first-class **Agent > Models** providers.
Use [Ollama](/providers/ollama) for native Ollama, [LM Studio](/providers/lmstudio)
for localhost:1234, [vLLM-compatible](/providers/vllm) for vLLM, SGLang, TGI,
LocalAI, FastChat, and similar OpenAI-compatible local servers, and
[Custom Provider](/providers/custom) when the shortcut does not fit.

Provider setup shows account/endpoint health: reachable, auth ok, models
discovered, and private network approved. Token and cost accounting remains on
Usage, where it is based on Fased's local usage history rather than provider
quota screens.