# AI

Source: https://athas.dev/docs/ai

Understand the AI features in Athas and choose the right workflow.

Athas has several AI surfaces. They share provider and model settings, but they are built for
different jobs.

## Pick the Right Tool

| Surface                                    | Use it for                                                                     |
| ------------------------------------------ | ------------------------------------------------------------------------------ |
| [Athas Agent](https://athas.dev/docs/athas-agent)           | Chat with a model, ask questions about code, and work with project context     |
| [Inline AI](https://athas.dev/docs/inline-ai)               | Get completions while typing or rewrite selected code in place                 |
| [External Agents](https://athas.dev/docs/external-agents)   | Run CLI coding agents published in the Integrations catalog                    |
| [Cloud Agents](https://athas.dev/docs/cloud-agents)         | Run Claude Code or Codex in a cloud sandbox and get a draft pull request       |
| [Providers and Models](https://athas.dev/docs/ai-providers) | Configure API keys, Ollama, OpenRouter, and custom OpenAI-compatible endpoints |

## How Configuration Works

AI settings live in **Settings > AI**.

- A default model is shared by new chats and other AI features unless overridden
- Other features can select their own model under **Settings > AI > Other features**
- Tab completion has a separate model picker
- Custom OpenAI-compatible endpoints can be used for either Athas Agent or autocomplete
- API keys are stored locally in secure storage
- Local chat history can be cleared from the same settings page

## Context and tools

Use [Agent context](https://athas.dev/docs/agent-context) to configure project rules and excluded paths.
[MCP servers](https://athas.dev/docs/mcp) provide additional tools; their transport support depends on the agent.

## Providers

Athas supports built-in providers, Ollama, OpenRouter, and custom OpenAI-compatible endpoints.

For setup details, see [Providers and Models](https://athas.dev/docs/ai-providers).

## Hosted AI Pricing

Plus includes $10 of Athas credit every month and Pro includes $20, on both monthly and
yearly billing. Team adds $40 of pooled credit per seat every month. Subscriptions started on
the original $10 Pro plan are now Plus.
Hosted autocomplete, inline edits, commit messages, drafts, integration generation, and
hosted Agent models share this credit. It refreshes monthly and unused credit does not
roll over.

Every hosted request costs the model's API list price plus 35%, for the included credit, the
Team pool, and any balance you add. When the included credit runs out, hosted models continue from a
prepaid balance you top up with $10, $20, $50 or $100, up to a monthly limit you set. A paid plan
is not required to use a prepaid balance. Your billing settings show each model's list price and the rate
you pay, plus available, used, and pending amounts; pending amounts settle once usage is
confirmed.

Most hosted models, and Automatic, read images: paste a PNG, JPEG, GIF or WebP screenshot of
up to 1.5 MB into the Agent composer. When the selected model reads text only, Athas leaves
the image out and says so.

Personal provider keys and local models never use the credit or the balance. Browser
usage is billed separately. [Cloud Agents](https://athas.dev/docs/cloud-agents) sandbox time and model usage are
paid from the same credit and balance.

## Choose the next step

Configure a connection in [Providers and Models](https://athas.dev/docs/ai-providers), give it [project context](https://athas.dev/docs/agent-context), then choose [Athas Agent](https://athas.dev/docs/athas-agent) for a local task or [Cloud Agents](https://athas.dev/docs/cloud-agents) for a repository sandbox.