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

# Connect ScitiX Docs to AI

> Learn how to use MCP, llms.txt, skill.md, and Markdown pages to let AI tools query Model Inference documentation.

You can connect ScitiX Docs to AI tools such as Claude Code, Codex, and Cursor. The AI tool can then answer Model Inference questions based on the documentation, such as how to choose a model, call the API, configure API keys, view usage, or connect development tools.

## Choose a method

We recommend using MCP first. MCP is best for querying ScitiX documentation continuously from an IDE, coding agent, or desktop AI client. The AI tool can retrieve relevant documentation on demand, without requiring you to copy page links or page content each time.

If your AI tool does not support MCP, choose another method based on your use case.

| Use case                                                         | Recommended method       |
| ---------------------------------------------------------------- | ------------------------ |
| Your AI tool supports MCP and needs long-term access to the docs | Use MCP                  |
| Your AI agent supports skills or custom context                  | Use `skill.md`           |
| You want AI to discover available documentation pages            | Use `llms.txt`           |
| You want AI to read the full documentation content               | Use `llms-full.txt`      |
| You only need AI to read one page                                | Use single-page Markdown |

## MCP access

If your AI tool supports Model Context Protocol (MCP), you can connect it to ScitiX documentation. After connection, the AI tool can search relevant documentation and answer questions based on the docs.

### Access URL

Use the following MCP server URL when adding ScitiX Docs to your AI tool.

```text theme={null}
https://docs.scitix.ai/mcp
```

### Tool capabilities

After connection, the AI tool can use the following documentation tools:

| Tool                    | Description                                                                                                                             |
| ----------------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
| `search`                | Search ScitiX documentation for relevant pages, guides, and API references.                                                             |
| `query_docs_filesystem` | Read pages from the virtual documentation filesystem. Use it to inspect full pages, browse the docs structure, or match exact keywords. |
| `submit_feedback`       | Submit feedback to the documentation team for incorrect, outdated, confusing, or incomplete documentation.                              |

### Example questions

* Which model should I use?
* How do I create an API key and call a model?
* How do I troubleshoot API errors?
* How do I view token usage and costs?

### Configure in Claude Code

Claude Code is primarily a command-line workflow. In the project directory where you want to use ScitiX Docs, run:

```bash theme={null}
claude mcp add --transport http scitix-docs https://docs.scitix.ai/mcp
```

After configuration, run the following command to check the connection:

```bash theme={null}
claude mcp list
```

### Configure in Codex

If you use a Codex client that supports graphical MCP configuration, you can add an MCP server in the app and set the server URL to `https://docs.scitix.ai/mcp`.

If you want to reuse the configuration across projects or IDEs, use the CLI or configuration file.

Run the following command to add the ScitiX Docs MCP server:

```bash theme={null}
codex mcp add scitixDocs --url https://docs.scitix.ai/mcp
```

After configuration, run the following command to check the connection:

```bash theme={null}
codex mcp list
```

You can also edit the Codex configuration file, such as `~/.codex/config.toml`, and add:

```toml theme={null}
[mcp_servers.scitixDocs]
url = "https://docs.scitix.ai/mcp"
```

## Other reading methods

### Use skill.md

`skill.md` describes the documentation site and product capabilities to AI agents. It is useful for tools that support skills or agent context.

```text theme={null}
https://docs.scitix.ai/skill.md
```

### Use llms.txt

`llms.txt` provides an index of the documentation pages. It is useful when you want AI to understand what pages are available.

```text theme={null}
https://docs.scitix.ai/llms.txt
```

### Use llms-full.txt

`llms-full.txt` provides more complete documentation content. It is useful when an AI tool needs a larger documentation context at once.

```text theme={null}
https://docs.scitix.ai/llms-full.txt
```

<Warning>
  `llms-full.txt` may contain a large amount of content. Some AI tools have context length limits. If the content is too long, use MCP, `llms.txt`, or single-page Markdown instead.
</Warning>

### Use single-page Markdown

Every documentation page can be read in Markdown format. Add `.md` to the page URL to get a plain-text version that is easier for AI tools to read.

For example:

```text theme={null}
https://docs.scitix.ai/model-inference/get-started/first-api-call.md
```

## Usage recommendations

* For Model Inference concepts, setup steps, and configuration details, prefer MCP or a specific Markdown page.
* For model availability, pricing, quotas, and console status, also check the console.
* If an AI answer conflicts with the documentation page, rely on the documentation page.