# OpenAI API

The OpenAI Responses API can connect directly to remote MCP servers. Pass the
Probo endpoint and OAuth token as an MCP tool; do not translate MCP tools into
custom REST routes.

:::caution
The model can access the Probo tools allowed by your token and request. Restrict the tool list and require approval for any operation that can change compliance data.
:::

## Prerequisites

- An OpenAI API key
- Access to a Probo instance
- A scoped Probo OAuth token
- A recent version of the OpenAI Python or JavaScript SDK

## Endpoint

Choose the endpoint for your Probo environment:

| Environment | MCP URL                                      |
| ----------- | -------------------------------------------- |
| Probo US    | `https://us.probo.com/api/mcp/v1`            |
| Probo EU    | `https://eu.probo.com/api/mcp/v1`            |
| Custom      | `https://your-probo-instance.com/api/mcp/v1` |

The `/v1` segment is required.

## Example

Install the SDK and set both credentials:

  

```bash
pip install --upgrade openai
export OPENAI_API_KEY="your_openai_api_key"
export PROBO_API_TOKEN="your_probo_api_token"
```

  
  

```bash
npm install openai
export OPENAI_API_KEY="your_openai_api_key"
export PROBO_API_TOKEN="your_probo_api_token"
```

  

Call a read-only Probo tool through the Responses API:

  

```python
from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-5.6",
    input="List the Probo organizations I can access.",
    tools=[
        {
            "type": "mcp",
            "server_label": "probo",
            "server_description": "Probo compliance management",
            "server_url": "https://us.probo.com/api/mcp/v1",
            "authorization": os.environ["PROBO_API_TOKEN"],
            "allowed_tools": ["listOrganizations"],
            "require_approval": "never",
        }
    ],
)

print(response.output_text)
```

  
  

```javascript

const client = new OpenAI();

const response = await client.responses.create({
  model: "gpt-5.6",
  input: "List the Probo organizations I can access.",
  tools: [
    {
      type: "mcp",
      server_label: "probo",
      server_description: "Probo compliance management",
      server_url: "https://us.probo.com/api/mcp/v1",
      authorization: process.env.PROBO_API_TOKEN,
      allowed_tools: ["listOrganizations"],
      require_approval: "never",
    },
  ],
});

console.log(response.output_text);
```

  

This request skips approval only because `allowed_tools` contains one read-only operation.

## Approval policy

Omit `require_approval` or set it to `"always"` when the model can use write operations. The response will contain an MCP approval request that your application must present to a user and send back in a follow-up Responses API call.

For lower latency, you can skip approval for an explicit set of read-only tools:

```json
{
  "allowed_tools": ["listOrganizations", "listRisks", "getRisk"],
  "require_approval": {
    "never": {
      "tool_names": ["listOrganizations", "listRisks", "getRisk"]
    }
  }
}
```

Do not use `"require_approval": "never"` without also restricting `allowed_tools` when the token can create or update data.

## Authentication behavior

The `authorization` value is your Probo OAuth token. OpenAI sends it to the MCP
server for that request but does not store it in the Response object. Include
it in every Responses API request that configures the Probo MCP tool.

OAuth access is limited by both the token's scopes and the Probo permissions of
the user who created it. See
[Authentication](/docs/developers/api/mcp/authentication) for scope, expiration,
and rotation guidance.

## Troubleshooting

### The MCP server returns 401

- Confirm `PROBO_API_TOKEN` is set in the application process.
- Confirm the token is active and has access to the requested organization.
- Include `authorization` in every Responses API request.

### The MCP server returns 404

Confirm that `server_url` ends in `/api/mcp/v1`. The unversioned `/api/mcp` route and REST-style `/tools/{name}` routes are not valid Probo MCP endpoints.

### The model cannot call a tool

- Confirm the exact tool name is present in `allowed_tools`.
- Confirm the tool is available in the [Probo MCP reference](/docs/developers/api/mcp/tools).
- Test the same endpoint with the [MCP Inspector](/docs/developers/api/mcp/overview#server-exploration).

For the Responses API approval flow and complete MCP tool schema, see the [OpenAI MCP documentation](https://developers.openai.com/api/docs/guides/tools-connectors-mcp).