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

# Agent Access

> All AIHubMix access entry points for AI agents: the agents.md onboarding guide, the llms.txt machine-readable index, per-model llms.txt call guides, the Playground Skill, and Playground deep links. One copyable command per entry point.

AIHubMix provides a set of machine-readable entry points for AI agents (coding assistants such as Claude Code, Codex, Cursor, and Cline). An agent can discover models, generate calls, verify connectivity, and hand back a preconfigured Playground link on its own. This page lists every entry point, each with one copyable command.

## One-line start

Send this sentence to your AI agent:

```text theme={null}
Fetch https://aihubmix.com/agents.md and follow it.
```

`agents.md` is an onboarding guide written for agents to execute directly: check the API key, pick a model from the live catalog, call it by protocol, verify end to end, and consult the included troubleshooting table.

## Entry point overview

| Entry point                 | URL                                                         | Purpose                                                                                       |
| --------------------------- | ----------------------------------------------------------- | --------------------------------------------------------------------------------------------- |
| Agent onboarding guide      | `https://aihubmix.com/agents.md`                            | Step-by-step instructions for agents to complete onboarding and verification on their own     |
| Site machine-readable index | `https://aihubmix.com/llms.txt`                             | LLM-readable index of site-wide capabilities and documentation                                |
| Docs machine-readable index | `https://docs.aihubmix.com/llms.txt`                        | Full page index of the docs site; append `.md` to any docs page URL to get plain text         |
| Per-model call guide        | `https://aihubmix.com/model/{model_id}/llms.txt`            | Pricing, verified capabilities, protocol endpoints, and runnable examples for a single model  |
| Playground Skill            | `https://aihubmix.com/skills/playground`                    | Agent skill file for trying models, tuning parameters, comparing models, and generating media |
| Live model catalog          | `https://aihubmix.com/api/v1/models?type=llm&sort_by=order` | Auth-free JSON with pricing and metadata                                                      |

## Per-model llms.txt

Every model has a machine-readable call guide covering pricing, a verified capability list, supported protocol endpoints, and runnable examples:

```bash theme={null}
curl -s https://aihubmix.com/model/gemini-3.7-flash/llms.txt
```

The `llms.txt` tab on a model detail page opens the corresponding file directly. Models that have not completed verification return a guidance document pointing to the live catalog and parameter schema, with no unverified capability claims.

## Playground Skill

Hand the Playground's capabilities (model trials, parameter tuning, side-by-side comparison, image and video generation, deep-link handoff) to your agent:

```text theme={null}
Run curl -fsSL https://aihubmix.com/skills/playground and follow it.
```

The skill is also published at `/.well-known/skills/aihubmix-playground/SKILL.md`, and can be installed from the [GitHub repository](https://github.com/AIhubmix/skills) with `npx skills add AIhubmix/skills`. For onboarding and integration scenarios, use the [AIHubMix API Skill](/en/skills).

## Playground deep links

Agents can generate direct links so you can take over a preconfigured Playground session in your browser:

* Multi-model comparison: `https://playground.aihubmix.com/?models=gemini-3.7-flash,claude-opus-4-1` (up to 6 models)
* Direct configuration: `https://playground.aihubmix.com/?config=<base64url-encoded JSON>`, where the JSON takes the form `{"model": "...", "proto": "chat", "params": {...}, "system": "...", "draft": "..."}`; `draft` prefills the input box without sending

Link parameters are single-use: the URL parameters are cleared once the page loads, and no keys are carried in the link.

## Code generation

The npm package [`@aihubmix/codegen`](https://www.npmjs.com/package/@aihubmix/codegen) provides programmatic code generation for 4 protocols and 7 languages, and the request bodies it constructs match the requests the Playground actually sends. [`@aihubmix/model-schema`](https://www.npmjs.com/package/@aihubmix/model-schema) converts parameter schemas into codegen input.

## FAQ

### How does an agent verify a successful integration?

`agents.md` includes end-to-end verification steps: send a minimal request that demands an exact reply, then check the response body and the echoed model. HTTP 200 does not equal success; the response content must be checked.

### What credentials are required?

One AIHubMix API key, passed through the `AIHUBMIX_API_KEY` environment variable and created in the [console](https://console.aihubmix.com/token). Do not write the key into code, scripts, or conversations.

### Which models does per-model llms.txt cover?

Verified models return full documentation (the capability list comes from officially verified data), and the remaining models return a guidance document. Coverage expands as verification progresses.

### What is the current state of MCP support?

The MCP endpoint currently serves image generation, with capability expansion in progress. At this stage, the Skill and agents.md are the recommended entry points.

***

Last updated: 2026-08-20