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

> ## Agent Instructions
> Tokios exposes private local models through https://api.tokios.com using an outbound-only connector tunnel.
> Use 'connector' for the local tokios-connector process.
> Use 'deployment' for the public model name in the API model field.
> Use 'upstream model id' for the model served by Ollama, llama.cpp, vLLM, LM Studio, or another local backend.
> API keys use the prefix sk-tok-. Connector tokens use the prefix ct-tok-.

# Use your local model from your phone

> Keep your model running at home and reach it from an iOS or Android chat app by pointing the app at your Tokios endpoint with an sk-tok- API key.

Your model stays exactly where it is — on the machine running Ollama, llama.cpp, vLLM, or LM Studio. To reach it from your phone, install a chat app that supports a custom OpenAI-compatible endpoint, and point it at `https://api.tokios.com/v1` with your `sk-tok-…` API key. No VPN, no port forwarding, no exposing your home network.

This page covers the OpenAI-compatible surface only. Few mobile chat apps expose a settable Anthropic base URL, so treat `ANTHROPIC_BASE_URL` as a desktop/CLI feature (see [Use Claude Code](https://docs.tokios.com/guides/claude-code.md)) unless you find one that does.

## The one gotcha that breaks every setup: `/v1` or not

Mobile apps disagree on whether the base URL field wants the `/v1` suffix:

* **Apps that want `/v1` included:** Maid, Open WebUI, LibreChat. Use `https://api.tokios.com/v1`.
* **Apps that want the bare host and append the path themselves:** Chatbox. Use `https://api.tokios.com` (no `/v1`).

Get this backwards and requests 404 or never leave the app. Check the field's placeholder text — if it shows a default ending in `/v1` (for example `https://api.openai.com/v1`), include `/v1`. If the field is just labeled **API Host**, leave `/v1` off.

## Apps to avoid

* **Chatbox on iOS** — do not use. A confirmed, unresolved bug ([chatboxai/chatbox#3516](https://github.com/chatboxai/chatbox/issues/3516)) means custom OpenAI-compatible endpoints with Bearer-key auth fail on iOS: non-proxied requests go through a WKWebView `fetch()` that enforces CORS, and the `Authorization` header triggers a preflight `OPTIONS` request that Tokios (like most local-facing servers) doesn't answer the way the app expects. Chatbox on **Android** does not have this bug — see below.
* **Reins** and **Enchanted** — do not use. Both apps speak only the Ollama-native API (`/api/chat`, `/api/generate`). Tokios exposes the OpenAI-compatible (`/v1/chat/completions`) and Anthropic-compatible (`/v1/messages`) surfaces, not the raw Ollama API, so these apps cannot connect at all unless a future release adds an OpenAI-compatible mode.

> [tab] iOS
>
> All three apps below take a custom endpoint and a Bearer API key. None of them are Tokios-specific, so confirm the exact field labels in-app before you rely on them — they can change between app versions.
>
> ### Pal Chat
>
> Supports custom OpenAI-compatible endpoints and bring-your-own-key providers.
>
> | Field | Value |
> | - | - |
> | Provider type | Custom / OpenAI-compatible provider — the exact menu label varies by version, so check yours |
> | Base URL / Endpoint | `https://api.tokios.com/v1` — confirm whether Pal wants `/v1` or the bare host in your version |
> | API Key | `sk-tok-YOUR_KEY` |
> | Model | your Tokios deployment name |
>
> ### Apollo (by Liquid AI)
>
> Connects to local or custom AI servers that use the OpenAI-compatible format, and lets you specify a model identifier for custom backends.
>
> | Field | Value |
> | - | - |
> | Add provider → Custom / OpenAI-compatible endpoint — exact label varies by version | — |
> | Base URL | `https://api.tokios.com/v1` — confirm `/v1` handling against your version |
> | API Key | `sk-tok-YOUR_KEY` |
> | Model | your Tokios deployment name |
>
> ### 3sparks Chat
>
> Describes itself as compatible with OpenAI-compatible APIs, local and online. Exact field labels are unconfirmed.
>
> | Field | Value |
> | - | - |
> | Provider | OpenAI-compatible / Custom — exact label varies by version |
> | Base URL | `https://api.tokios.com/v1` — confirm `/v1` handling against your version |
> | API Key | `sk-tok-YOUR_KEY` |
> | Model | your Tokios deployment name |
>
> > [!WARNING]
> >
> > Do not use **Chatbox** on iOS for this — see [Apps to avoid](#apps-to-avoid) above.

> [tab] Android
>
> ### Maid
>
> Open-source. Use its **OpenAI** provider (not the separate Ollama provider, which is a different code path and not for Tokios).
>
> | Field | Value |
> | - | - |
> | Provider | **OpenAI** |
> | Base URL | `https://api.tokios.com/v1` — include `/v1` (Maid's own default is `https://api.openai.com/v1`) |
> | API Key | `sk-tok-YOUR_KEY` |
> | Model | select from the **Model** dropdown |
> | Custom Headers | leave empty |
>
> > [!WARNING]
> >
> > Maid's **Model** field is a dropdown populated by calling the endpoint's model list (`GET /v1/models`), not a free-text box. It only shows deployments that Tokios returns for your API key, so register the deployment and confirm it appears in `GET /v1/models` for that key before expecting to pick it in Maid. If the dropdown is empty, check that the key is scoped to the deployment's name pattern on the **Keys** tab.
>
> Maid also lists Anthropic as a remote provider, but it's not clear whether it exposes a *custom* Anthropic base URL — check the current version before relying on it. Use the OpenAI provider above for Tokios.
>
> ### Chatbox
>
> Works fine on Android — the iOS bug above is iOS-specific.
>
> | Field | Value |
> | - | - |
> | Add Custom Provider → API Mode | **OpenAI API Compatible** |
> | API Host | `https://api.tokios.com` — **no `/v1`**; Chatbox appends `/v1/chat/completions` itself |
> | API Path | leave empty |
> | API Key | `sk-tok-YOUR_KEY` |
> | Model | type your Tokios deployment name manually if it doesn't auto-list |
>
> > [!TIP]
> >
> > Maid wants `/v1` in the base URL; Chatbox wants the bare host. Double-check which field you're filling in before you copy a URL between apps.

## Power option: run a hosted web UI and install it as a PWA

If you want a fuller chat interface (chat history, multiple deployments, system prompts) instead of a dedicated native app, host **Open WebUI** or **LibreChat** on a server, point its OpenAI-compatible connection at Tokios, and install it on your phone as a Progressive Web App:

**1. Configure Tokios on the server**

In Open WebUI: **Admin → Settings → Connections → OpenAI API**. Set:

| Field | Value |
| - | - |
| API Base URL | `https://api.tokios.com/v1` |
| API Key | `sk-tok-YOUR_KEY` |
| Model | your Tokios deployment name (appears in the model picker) |

LibreChat's equivalent is its OpenAI-compatible custom endpoint configuration on the server side.

**2. Install the PWA on your phone**

Open the hosted instance's URL in your phone's browser, then:

* **iOS (Safari only):** tap **Share → Add to Home Screen**.
* **Android (Chrome, Firefox, Brave, Edge):** open the browser menu → **Install app** / **Add to Home Screen**.

The Tokios configuration lives entirely on the server — your phone never holds the API key directly; it just opens the installed PWA against your hosted Open WebUI or LibreChat instance.

## Next steps

> [card] [Quickstart](https://docs.tokios.com/quickstart.md)
>
> Go from zero to a working connector, deployment, and API key.

> [card] [Add API keys to local models](https://docs.tokios.com/guides/add-api-keys-to-local-models.md)
>
> Scope a key to specific deployments before you hand it to a mobile app.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.