# Claude Desktop 3P Mode with LiteLLM

Claude Desktop has a **third-party inference (3P) mode** that allows the Desktop application to route inference requests through a custom gateway such as LiteLLM. This is configured separately from the normal Anthropic account login. Recent Claude Desktop configurations use a `Claude-3p` configuration directory and an `inferenceProvider: "gateway"` configuration. ([Gist][1])

> **Important:** When configuring the LiteLLM gateway URL in Claude Desktop 3P mode, **do not include `/v1` in the Base URL**. Claude Desktop adds the `/v1/...` API paths itself. For example, use `http://localhost:4000`, **not** `http://localhost:4000/v1`. A working 3P gateway will ultimately receive requests such as `/v1/models` and `/v1/messages`. ([GitHub][2])

## 1. Enable Developer Mode

Open Claude Desktop and go to:

**Help → Troubleshooting → Enable Developer Mode**

After enabling Developer Mode, Claude Desktop exposes the developer configuration options, including:

**Developer → Configure third-party inference**

This is the preferred way to create the 3P configuration.

## 2. Configure the LiteLLM gateway

In **Configure third-party inference**, select the gateway/custom provider option.

Use your LiteLLM proxy as the gateway.

For a local LiteLLM installation:

```text
Gateway/Base URL:
http://localhost:4000

API Key:
sk-your-litellm-key
```

For a remote LiteLLM installation:

```text
Gateway/Base URL:
https://litellm.example.com

API Key:
sk-your-litellm-key
```

### Do NOT use this

```text
http://localhost:4000/v1
```

### Use this instead

```text
http://localhost:4000
```

LiteLLM itself exposes the Anthropic-compatible API under paths such as:

```text
/v1/messages
/v1/models
```

Claude Desktop's gateway configuration supplies those API paths. ([GitHub][2])

## 3. LiteLLM configuration

For example, a minimal LiteLLM configuration could look like:

```yaml
model_list:
  - model_name: claude-sonnet
    litellm_params:
      model: anthropic/claude-sonnet-4-6
      api_key: os.environ/ANTHROPIC_API_KEY

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY
```

Start LiteLLM on port 4000:

```bash
litellm --config config.yaml --port 4000
```

Then Claude Desktop should connect to:

```text
http://localhost:4000
```

while LiteLLM handles:

```text
http://localhost:4000/v1/messages
http://localhost:4000/v1/models
```

## 4. What the resulting 3P configuration looks like

Claude Desktop stores its 3P configuration separately from the normal Desktop configuration. On macOS, for example, configurations can be found under:

```text
~/Library/Application Support/Claude-3p/
```

A representative configuration contains:

```json
{
  "deploymentMode": "3p",
  "enterpriseConfig": {
    "inferenceProvider": "gateway",
    "inferenceGatewayBaseUrl": "http://localhost:4000",
    "inferenceGatewayApiKey": "sk-your-litellm-key",
    "inferenceGatewayAuthScheme": "bearer",
    "inferenceModels": [
      "claude-sonnet-4-6"
    ]
  }
}
```

The exact configuration structure can vary between Claude Desktop releases, so it is preferable to let **Configure third-party inference** generate the configuration rather than manually replacing the entire file. Current examples confirm the `deploymentMode: "3p"` and `inferenceProvider: "gateway"` structure. ([Gist][1])

## 5. Restart Claude Desktop

After saving the third-party inference configuration, allow Claude Desktop to restart.

You should now be running in **3P mode**, with inference routed approximately as follows:

```text
┌──────────────────────┐
│    Claude Desktop    │
│                      │
│       3P Mode        │
└──────────┬───────────┘
           │
           │ Base URL:
           │ http://localhost:4000
           ▼
┌──────────────────────┐
│       LiteLLM        │
│                      │
│     /v1/messages     │
│      /v1/models      │
└──────────┬───────────┘
           │
           ▼
┌──────────────────────┐
│   Configured model   │
│                      │
│ Anthropic / OpenAI / │
│ Bedrock / etc.       │
└──────────────────────┘
```

## 6. Verify LiteLLM independently

Before troubleshooting Claude Desktop, verify that LiteLLM is responding.

For example:

```bash
curl http://localhost:4000/v1/models \
  -H "Authorization: Bearer sk-your-litellm-key"
```

You should get a model list.

You can also test the Anthropic Messages endpoint directly:

```bash
curl http://localhost:4000/v1/messages \
  -H "x-api-key: sk-your-litellm-key" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet",
    "max_tokens": 100,
    "messages": [
      {
        "role": "user",
        "content": "Say hello"
      }
    ]
  }'
```

If these work but Claude Desktop doesn't, the problem is likely the 3P configuration rather than LiteLLM.

## 7. Authentication caveat

There is an important distinction between **authenticating to LiteLLM** and **authenticating LiteLLM to Anthropic**.

Your Claude Desktop 3P API key:

```text
sk-your-litellm-key
```

authenticates **Claude Desktop → LiteLLM**.

If LiteLLM is routing to Anthropic, LiteLLM still needs valid credentials for **LiteLLM → Anthropic**. Current testing indicates that Claude Desktop's 3P/Cowork path does not necessarily forward an existing Claude subscription OAuth credential to LiteLLM, so a normal Anthropic API key may be required on the LiteLLM side. ([GitHub][3])

For example:

```yaml
model_list:
  - model_name: claude-sonnet
    litellm_params:
      model: anthropic/claude-sonnet-4-6
      api_key: os.environ/ANTHROPIC_API_KEY
```

This is different from the Claude Code CLI, which has a different credential-forwarding behavior. ([GitHub][3])

## Quick reference

| Setting             | Value                                              |
| ------------------- | -------------------------------------------------- |
| Mode                | `3p`                                               |
| Provider            | `gateway`                                          |
| LiteLLM local URL   | `http://localhost:4000`                            |
| **Include `/v1`?**  | **No**                                             |
| API key             | Your LiteLLM virtual/master key                    |
| LiteLLM API paths   | `/v1/messages`, `/v1/models`                       |
| LiteLLM → Anthropic | Requires appropriate Anthropic credentials         |
| Developer Mode      | **Help → Troubleshooting → Enable Developer Mode** |
| Configuration       | **Developer → Configure third-party inference**    |

**The key gotcha:** **Claude Desktop 3P Base URL = the gateway root, not the API version path.** So if LiteLLM is listening on port 4000, enter `http://localhost:4000`, not `http://localhost:4000/v1`. ([GitHub][2])

[1]: https://gist.github.com/avarayr/a9a35354aa6d7d8430ce0c27cd9aff3f?permalink_comment_id=6137995&utm_source=chatgpt.com "~/Library/Application Support/Claude-3p/claude_desktop_config.json · GitHub"
[2]: https://github.com/farion1231/cc-switch/issues/2982?utm_source=chatgpt.com "Claude Desktop 3P mode: WebFetch fails because domain safety check is not allowed/proxied · Issue #2982 · farion1231/cc-switch · GitHub"
[3]: https://github.com/BerriAI/litellm/discussions/30827?utm_source=chatgpt.com "Claude Chat/Cowork Desktop App LiteLLM Integration · BerriAI litellm · Discussion #30827 · GitHub"