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

# OpenCode

## Connect a third-party API in OpenCode

OpenCode is an open-source coding assistant with flexible API wiring. Point it at a relay or direct provider for better models or pricing (DeepSeek, SiliconFlow, etc.).

***

## Step 1: Gather credentials

1. **API key**: [Tokensmind console](https://tokensmind.ai/token)
2. **Base URL**: e.g. `https://tokensmind.ai/v1`
3. **Model id**: e.g. `deepseek-ai/DeepSeek-V3`

***

## Step 2: Open settings

1. Launch **OpenCode**.
2. **Settings gear** (lower left) → **Settings**.
3. Search for `OpenCode` or `AI Provider`.
4. Set **Provider** to `OpenAI Compatible` (or `Custom`).

***

## Step 3: Fill in fields

### 1. Base URL

Set **OpenAI Compatible: Base URL** to your proxy.

> End with `/v1`, **without** `/chat/completions`.

### 2. API key

Paste into **OpenAI Compatible: API Key**.

### 3. Models

Under **Custom Models** / **Model Map**:

* **Add Item**
* Display name (e.g. `DeepSeek`)
* Real model id from the provider (e.g. `deepseek-chat`)

***

## Step 4: Test

1. Open **Chat** (robot icon in the sidebar).
2. Pick your **custom model** in the switcher.
3. Send a prompt (e.g. “implement quicksort”) to verify connectivity.

***

## Advanced: environment variables

OpenCode can read `.env` or system env:

* `OPENAI_API_BASE` — third-party base URL
* `OPENAI_API_KEY` — key

***

## Troubleshooting

| Error                 | Likely cause          | Fix                                 |
| :-------------------- | :-------------------- | :---------------------------------- |
| **Model Not Found**   | Wrong model id        | Match the provider list exactly.    |
| **Invalid API Key**   | Bad or expired key    | Regenerate; strip newlines.         |
| **Network Error**     | TLS / proxy           | Adjust proxy for CN relays.         |
| **Stream Mode Error** | Streaming unsupported | Try disabling **Enable Streaming**. |

***

> **Tip**: OpenCode works well with local stacks. For **LM Studio** or **LocalAI**, set Base URL to `http://localhost:xxxx/v1` for offline use.