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

# Provider Configuration

> Setup guides for each LLM provider

## Anthropic (Claude)

```bash theme={null}
export ANTHROPIC_BASE_URL=https://proxy.meshai.dev/v1/anthropic/k/msh_YOUR_KEY
```

Works with: Claude Code, Claude SDK, any tool using `ANTHROPIC_BASE_URL`.

## OpenAI (GPT, o1)

```bash theme={null}
export OPENAI_BASE_URL=https://proxy.meshai.dev/v1/openai/k/msh_YOUR_KEY
```

Works with: Codex, Cline, Cursor, any tool using `OPENAI_BASE_URL`.

## NVIDIA (Nemotron)

NVIDIA's API is OpenAI-compatible, so there is no `NVIDIA_BASE_URL` env var. Point an OpenAI-compatible client at the proxy instead:

```python theme={null}
from openai import OpenAI

client = OpenAI(
    base_url="https://proxy.meshai.dev/v1/nvidia/k/msh_YOUR_KEY",
    api_key="your-nvidia-api-key",
)
```

## Google Vertex AI

Vertex is configured on the client, not through an env var, and requires the location and project as part of the path:

```
https://proxy.meshai.dev/v1/vertex/k/msh_YOUR_KEY/{location}/{project}/publishers/google/models/{model}:generateContent
```

Example: `/v1/vertex/us-central1/my-project/publishers/google/models/gemini-2.0-flash:generateContent`

## AWS Bedrock

Bedrock is configured through the boto3 client's `endpoint_url`, not an env var, and the region must be part of the path:

```python theme={null}
import boto3

bedrock = boto3.client(
    "bedrock-runtime",
    endpoint_url="https://proxy.meshai.dev/v1/bedrock/k/msh_YOUR_KEY/us-east-1",
)
```

## Custom Providers

Contact us to add support for additional providers. The proxy architecture supports any HTTP-based LLM API.