### Development Setup Commands

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/README.md

Commands for setting up the development environment, installing dependencies, and running tests.

```bash
uv sync --all-extras --dev
uv run pre-commit install
uv run pytest -v
uv run ruff check src/ tests/
```

--------------------------------

### NewProjectRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for creating a new project.

```json
{
  "name": "Project Phoenix",
  "organization_id": "org-456"
}
```

--------------------------------

### Clone Repository and Install Dependencies

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/README.md

Clone the litellm-mcp repository and install its dependencies using uv.

```bash
git clone https://github.com/TETRA-2023/litellm-mcp.git
cd litellm-mcp
uv sync
```

--------------------------------

### NewUserRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for creating a new internal user.

```json
{
  "username": "newuser",
  "email": "newuser@example.com",
  "role": "editor"
}
```

--------------------------------

### NewCustomerRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for creating a new customer account.

```json
{
  "name": "New Customer Inc.",
  "email": "contact@newcustomer.com",
  "plan_id": "basic"
}
```

--------------------------------

### NewOrganizationRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for creating a new organization.

```json
{
  "name": "New Org LLC",
  "owner_email": "owner@neworg.com"
}
```

--------------------------------

### MCPServer Request Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON for creating a new MCP Gateway Server.

```json
{
  "name": "gateway-1",
  "url": "http://gateway.example.com",
  "description": "Primary gateway instance."
}
```

--------------------------------

### Project Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'Project' resource, showing its name and associated organization.

```json
{
  "id": "proj-abc",
  "name": "Alpha Project",
  "organization_id": "org-123",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### MCPTool Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an MCP Tool resource.

```json
{
  "id": "list_models",
  "name": "List Models",
  "description": "Lists all available models.",
  "parameters": {},
  "returns": {"type": "array", "items": {"type": "object"}}
}
```

--------------------------------

### Run the LiteLLM Server

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/README.md

Command to start the LiteLLM server using uv and Python.

```bash
uv run python src/server.py
```

--------------------------------

### Budget Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'Budget' resource, including name, amount, and currency.

```json
{
  "id": "budget-q3",
  "name": "Q3 Marketing Budget",
  "amount": 5000.0,
  "currency": "USD",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### NewMCPServerRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for registering a new MCP Gateway Server.

```json
{
  "name": "MCP-Server-01",
  "url": "http://mcp.example.com/api",
  "description": "Main MCP server instance."
}
```

--------------------------------

### User Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'User' resource, including username and email.

```json
{
  "id": "user-123",
  "username": "testuser",
  "email": "test@example.com",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### Customer Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'Customer' resource, showing name and contact information.

```json
{
  "id": "cust-xyz",
  "name": "Acme Corp",
  "email": "contact@acme.com",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z",
  "status": "active"
}
```

--------------------------------

### Model Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'Model' resource, showing its fields.

```json
{
  "id": "gpt-3.5-turbo",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z",
  "input_cost": 0.0000015,
  "output_cost": 0.000002,
  "description": "OpenAI's fast and affordable model."
}
```

--------------------------------

### Budget Object Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Example of a budget object with standard fields. Used for tracking spending limits and reset periods.

```json
{
    "budget_id": "budget-1",
    "max_budget": 10000.0,
    "soft_budget": 8000.0,
    "budget_duration": "monthly",
    "tpm_limit": 1000000,
    "rpm_limit": 5000,
    "model_max_budget": {
        "gpt-4o": 5000.0,
        "claude-3": 3000.0
    },
    "budget_reset_at": "2025-07-01T00:00:00Z"
}
```

--------------------------------

### Tool Invocation Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Demonstrates a basic pattern for invoking an MCP tool.

```python
tool_result = await client.mcp_gateway.invoke_tool(
    tool_name="list_models",
    tool_params={}
)
print(tool_result)
```

--------------------------------

### Configure Environment Variables

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/README.md

Copy the example environment file and edit it with your LiteLLM proxy URL and API key.

```bash
cp .env.example .env
# Edit .env with your LiteLLM proxy URL and API key
```

--------------------------------

### GenerateKeyRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for generating a new API key.

```json
{
  "name": "my-new-key",
  "scopes": ["read", "write"],
  "expires_in_days": 30
}
```

--------------------------------

### Credential Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'Credential' resource, including its type and name.

```json
{
  "id": "cred-123",
  "name": "openai-api-key",
  "type": "api_key",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### Key Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an API 'Key' resource, showing its ID, name, and status.

```json
{
  "id": "key-abc",
  "name": "app-key-1",
  "scopes": ["read", "write"],
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z",
  "status": "active"
}
```

--------------------------------

### SpendRecord Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of a 'SpendRecord' resource, detailing usage and cost.

```json
{
  "id": "spend-xyz",
  "resource_type": "chat_completion",
  "model_id": "gpt-3.5-turbo",
  "cost": 0.0005,
  "timestamp": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### Organization Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an 'Organization' resource, including its name and owner.

```json
{
  "id": "org-123",
  "name": "Tech Solutions Inc.",
  "owner_email": "admin@techsolutions.com",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### ChatCompletion Response Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON response for a chat completion request.

```json
{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1678886400,
  "model": "gpt-3.5-turbo",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello there! How can I help you today?"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 15,
    "total_tokens": 25
  }
}
```

--------------------------------

### AccessGroup Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an 'AccessGroup' resource, detailing its properties.

```json
{
  "id": "research-models",
  "name": "Research Models",
  "description": "Models available for research purposes.",
  "created_at": "2023-01-01T12:00:00Z",
  "updated_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### MCPSubmission Request Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON for submitting a tool execution request to the MCP Gateway.

```json
{
  "tool_id": "list_models",
  "parameters": {},
  "user_credential_id": "mcp-cred-123"
}
```

--------------------------------

### MCPServer Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents a Managed Compute Platform server. Use this structure when displaying standard server details.

```python
{
    "server_id": "server-123",
    "server_name": "Research Tools",
    "alias": "research-mcp",
    "description": "Internal research tool server",
    "transport": "http",
    "url": "http://localhost:8000",
    "auth_type": "bearer",
    "mcp_access_groups": ["ag-1"],
    "allowed_tools": ["search", "summarize"],
    "approval_status": "approved"
}
```

--------------------------------

### MCPCredential Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an MCP Credential resource.

```json
{
  "id": "mcp-cred-123",
  "name": "User API Key",
  "type": "api_key",
  "user_id": "user-123"
}
```

--------------------------------

### MCPHealth Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an MCP Health resource.

```json
{
  "service": "MCP Gateway",
  "status": "operational",
  "details": "All systems nominal."
}
```

--------------------------------

### Get Budget Settings

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the current budget settings for the account.

```python
get_budget_settings()
```

--------------------------------

### Invoke LiteLLM Tool via Python Client

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/README.md

Example of using the LiteLLMClient to list models. Requires importing LiteLLMClient and settings.

```python
from src.litellm_client import LiteLLMClient
from src.config import settings

client = LiteLLMClient(
    base_url=settings.proxy_url,
    api_key=settings.get_api_key_value()
)

models = await client.list_models()
print(models)
```

--------------------------------

### Listing Available Models

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example of calling the `list_models` method on `LiteLLMClient` to retrieve a list of available models.

```python
models = await client.models.list_models()
print(models)
```

--------------------------------

### MCPSubmission Data Structure Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON representation of an MCP Submission resource, likely for task execution.

```json
{
  "id": "sub-xyz",
  "tool_id": "list_models",
  "status": "pending",
  "created_at": "2023-01-01T12:00:00Z"
}
```

--------------------------------

### Getting Model Information

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Demonstrates how to fetch detailed information about a specific model using its ID.

```python
model_info = await client.models.get_model_info(model_id="gpt-3.5-turbo")
print(model_info)
```

--------------------------------

### SpendRecord Object Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Example of a spend record object with standard fields. Captures details of a single API request's cost and token usage.

```json
{
    "request_id": "req-xyz",
    "call_type": "chat",
    "api_key": "sk-…",
    "user": "user-123",
    "team_id": "team-1",
    "model": "gpt-4o",
    "total_tokens": 2048,
    "spend": 0.15,
    "startTime": "2025-06-14T10:00:00Z",
    "endTime": "2025-06-14T10:00:05Z"
}
```

--------------------------------

### ChatCompletion Object Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Example of a chat completion object adhering to OpenAI's format. Includes response ID, model, choices, and usage statistics.

```json
{
    "id": "chatcmpl-abc123",
    "object": "chat.completion",
    "created": 1686789200,
    "model": "gpt-4o",
    "choices": [
        {
            "index": 0,
            "message": {
                "role": "assistant",
                "content": "Hello! How can I help you?"
            },
            "finish_reason": "stop"
        }
    ],
    "usage": {
        "prompt_tokens": 10,
        "completion_tokens": 15,
        "total_tokens": 25
    }
}
```

--------------------------------

### Embedding Response Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON response for an embedding request.

```json
{
  "id": "embd-abc",
  "object": "embedding",
  "data": [
    {
      "index": 0,
      "embedding": [
        0.001, 0.002, -0.001, ...
      ]
    }
  ],
  "model": "text-embedding-ada-002",
  "usage": {
    "prompt_tokens": 5,
    "total_tokens": 5
  }
}
```

--------------------------------

### Updating an Existing Model

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example of updating the configuration or details of an existing model using its ID.

```python
update_data = {"input_cost": 0.00015}
await client.models.update_model(model_id="my-custom-model", data=update_data)
print("Model updated successfully.")
```

--------------------------------

### Initializing LiteLLMClient with Custom Settings

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Demonstrates initializing `LiteLLMClient` with a custom proxy URL and API key. Ensure your API key is kept secure.

```python
from litellm import LiteLLMClient

client = LiteLLMClient(proxy_url="http://localhost:4000", api_key="sk-12345")
# Use client for API calls...
```

--------------------------------

### MCPCredential Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents user-level credentials for an MCP server. Use this to check if a credential is set and if OAuth is enabled for a server.

```python
{
    "server_id": "server-123",
    "has_credential": true,
    "oauth_enabled": false,
    "expires_at": null
}
```

--------------------------------

### Generate Key with Extras Dictionary

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/README.md

Example of generating a key using an 'extras' dictionary for less common fields.

```python
# Extras dict for uncommon fields
key = await generate_key(
    key_alias="api-key-1",
    extras={"allow_all_keys": True, "custom_field": "value"}
)
```

--------------------------------

### GET /model/info

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Fetches administrative information about model deployments, either for all deployments or a specific one.

```APIDOC
## GET /model/info

### Description
Get admin-side model info for all deployments or one by id.

### Method
GET

### Endpoint
/model/info

### Parameters
#### Query Parameters
- **litellm_model_id** (str) - Optional - Internal deployment ID.

#### Request Body
None

### Response
#### Success Response (200)
`{"data": [...]}` (list of all deployments or single deployment)

### Status Codes
200, 401, 403
```

--------------------------------

### Tool Invocation for Spend Reporting

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example of invoking a tool to retrieve spend reporting data.

```python
spend_report = await client.mcp_gateway.invoke_tool(
    tool_name="get_spend_report",
    tool_params={"organization_id": "org-123", "period": "monthly"}
)
print(spend_report)
```

--------------------------------

### Health Response Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON response for a health check request.

```json
{
  "status": "ok",
  "checks": {
    "database": "healthy",
    "redis": "healthy"
  }
}
```

--------------------------------

### MCPHealth Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents the health status of an MCP server. Use this to monitor server health, latency, and any error messages.

```python
{
    "server_id": "server-123",
    "server_name": "Research Tools",
    "status": "healthy",
    "last_checked": "2025-06-14T10:05:00Z",
    "latency_ms": 45,
    "error_message": null
}
```

--------------------------------

### MCP Toolset Example (Standard Fields)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Demonstrates the structure of an MCP toolset with standard fields, including its ID, name, description, included tools, and creation/update timestamps. Useful for understanding the data returned by list_mcp_toolsets(), get_mcp_toolset(), add_mcp_toolset(), and update_mcp_toolset().

```python
{
    "toolset_id": "toolset-1",
    "toolset_name": "research-tools",
    "description": "Research and data analysis tools",
    "tools": [
        {"server_id": "server-123", "name": "search"},
        {"server_id": "server-456", "name": "summarize"}
    ],
    "created_at": "2025-06-01T00:00:00Z",
    "updated_at": "2025-06-14T10:00:00Z"
}
```

--------------------------------

### GET /v1/mcp/toolsets

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Lists all available MCP toolsets. Toolsets are collections of tools that can be managed together.

```APIDOC
## GET /v1/mcp/toolsets

### Description
List MCP toolsets.

### Method
GET

### Endpoint
/v1/mcp/toolsets

### Response
#### Success Response (200)
- **List of toolset dicts** - A list where each element is a dictionary representing an MCP toolset.

#### Response Example
```json
{
  "example": "response body"
}
```
```

--------------------------------

### Get Model Info

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves admin-side model information for all deployments or a specific one by its LiteLLM ID.

```python
all_models = await get_model_info()
one_model = await get_model_info(litellm_model_id="123e4567")
```

--------------------------------

### Create a Virtual Key

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/README.md

Example of creating a new virtual API key with specified models, budget, and user ID using the LiteLLM client.

```python
from src.litellm_client import LiteLLMClient

key = await client.generate_key(
    key_alias="prod-api-key",
    models=["gpt-4o", "claude-3-sonnet"],
    max_budget=1000.0,
    budget_duration="monthly",
    user_id="user-123"
)
print(f"New key: {key['token']}")
```

--------------------------------

### List Models with Minimal Response

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/README.md

Example of listing models with a minimal response verbosity. Can be called via the client or directly as an MCP tool.

```python
models = await client.list_models()  # Calls server.list_models()
# Or via MCP tool directly (in Claude)
# tool: list_models, params: {"verbosity": "minimal"}
```

--------------------------------

### UpdateKeyRequest Body Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example JSON payload for updating an existing API key.

```json
{
  "name": "my-updated-key",
  "scopes": ["read"]
}
```

--------------------------------

### Get Budget Information

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves budget information by a list of budget IDs. This is a batch lookup operation.

```python
get_budget_info(budget_ids=["budget_1", "budget_2"])
```

--------------------------------

### Get Public Model Hub Info

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Fetch metadata for the Model Hub, including title, description, and useful links.

```python
async def get_public_hub_info(self) -> dict:
```

```python
info = await client.get_public_hub_info()
print(info.get("title"))  # Model Hub title
```

--------------------------------

### GET /v1/mcp/tools

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Lists all available tools from registered MCP servers. This can be filtered by a specific server ID.

```APIDOC
## GET /v1/mcp/tools

### Description
List tools from registered MCP servers.

### Method
GET

### Endpoint
/v1/mcp/tools

### Parameters
#### Query Parameters
- **server_id** (str) - Optional - Filter tools by a specific MCP server ID.

### Response
#### Success Response (200)
- **List of MCP tool dicts** - A list where each element is a dictionary representing an MCP tool.

#### Response Example
```json
{
  "example": "response body"
}
```
```

--------------------------------

### Response Filtering Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Illustrates how to use response filtering to select specific fields from a tool's output. This is controlled by the `RESPONSE_FIELDS` mapping.

```python
# Assuming RESPONSE_FIELDS is configured elsewhere
filtered_response = await client.mcp_gateway.invoke_tool(
    tool_name="get_model_info",
    tool_params={"model_id": "gpt-3.5-turbo"},
    response_fields=["id", "input_cost"]
)
print(filtered_response)
```

--------------------------------

### Model Type Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents a model with its ID, object type, owner, and creation timestamp. Useful for listing and retrieving model information.

```python
{
    "id": "gpt-4o",
    "object": "model",
    "owned_by": "openai",
    "created": 1609459200
}
```

--------------------------------

### Tool Invocation for MCP Gateway Server Management

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Example of invoking a tool to manage MCP gateway servers, such as updating server details.

```python
server_update_result = await client.mcp_gateway.invoke_tool(
    tool_name="update_mcp_server",
    tool_params={"server_id": "gateway-1", "data": {"url": "http://new-gateway.example.com"}}
)
print(server_update_result)
```

--------------------------------

### Creating a New Budget

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Sets up a new budget, likely for tracking expenses.

```python
new_budget_data = {"name": "Q3 Marketing Budget", "amount": 5000.0, "currency": "USD"}
await client.budgets.create_budget(data=new_budget_data)
print("Budget created.")
```

--------------------------------

### Configuration using .env File

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/configuration.md

Sets LiteLLM MCP server configuration by creating a .env file in the project root. All variables are optional except LITELLM_API_KEY.

```bash
LITELLM_PROXY_URL=http://localhost:4000
LITELLM_API_KEY=sk-your-api-key
LITELLM_TRANSPORT=stdio
LITELLM_TIMEOUT_SECONDS=30
LITELLM_MAX_RETRIES=2
```

--------------------------------

### MCPTool Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents a tool available on an MCP server. Use this structure to display tool information, including its associated server and input schema.

```python
{
    "name": "search",
    "description": "Search the knowledge base",
    "server_id": "server-123",
    "inputSchema": {
        "type": "object",
        "properties": {
            "query": {
                "type": "string",
                "description": "Search query"
            }
        },
        "required": ["query"]
    }
}
```

--------------------------------

### GET /key/info

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Retrieves information about a specific virtual key. If no key is provided, it returns information about the caller's own key.

```APIDOC
## GET /key/info

### Description
Get info about a key.

### Method
GET

### Endpoint
/key/info

### Parameters
#### Query Parameters
- **key** (str) - Optional - The key to get information for. If omitted, returns the caller's own key.
- **verbosity** (str) - Optional - Verbosity level (defaults to 'standard').

### Response
#### Success Response (200)
- **Key info dict**

#### Error Response
- 401
- 403
- 404

### Note
Master key returns 404; test with a virtual key instead
```

--------------------------------

### Listing Credentials

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Fetches a list of configured credentials.

```python
credentials = await client.credentials.list_credentials()
print(credentials)
```

--------------------------------

### Initialize LiteLLMClient

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Initializes an async HTTP client for LiteLLM proxy operations. Requires base URL and API key. Supports optional timeout and retry settings.

```python
class LiteLLMClient:
    def __init__(
        self,
        base_url: str,
        api_key: str,
        timeout: float = 30.0,
        max_retries: int = 2,
    ) -> None:
        pass
```

```python
from src.litellm_client import LiteLLMClient
from src.config import settings

client = LiteLLMClient(
    base_url="http://localhost:4000",
    api_key="sk-1234567890",
    timeout=30.0,
    max_retries=2
)
```

--------------------------------

### Creating a New Customer

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Registers a new customer account.

```python
new_customer_data = {"name": "Acme Corp", "email": "contact@acme.com"}
await client.customers.create_customer(data=new_customer_data)
print("Customer created.")
```

--------------------------------

### User Type Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents a user with their ID, email, role, and associated teams/organizations. Includes spend and budget information. Standard verbosity omits some full fields.

```python
{
    "user_id": "user-123",
    "user_email": "alice@example.com",
    "user_alias": "alice",
    "user_role": "admin",
    "teams": ["team-1"],
    "organizations": ["org-1"],
    "models": [],
    "max_budget": null,
    "spend": 10.00,
    "blocked": false
}
```

--------------------------------

### Get MCP OpenAPI Registry

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the MCP OpenAPI registry. Corresponds to the GET /v1/mcp/openapi-registry API endpoint.

```python
get_mcp_openapi_registry() -> dict
```

--------------------------------

### LiteLLMClient Constructor

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Initializes an async HTTP client for LiteLLM proxy operations. Requires base URL and API key, with optional timeout and retry settings.

```APIDOC
## LiteLLMClient Constructor

### Description
Initializes an async HTTP client for LiteLLM proxy operations.

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Method
`__init__` (Constructor)

### Endpoint
N/A

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Parameters
| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| `base_url` | `str` | Yes | — | LiteLLM proxy base URL (e.g., `http://localhost:4000`); trailing `/` is stripped |
| `api_key` | `str` | Yes | — | Bearer token (master or virtual key); sent as `Authorization: Bearer <api_key>` |
| `timeout` | `float` | No | `30.0` | HTTP request timeout in seconds (passed to `httpx.AsyncClient`) |
| `max_retries` | `int` | No | `2` | Retry budget for transient errors (408, 429, 500–504); 0 means no retry |

### Request Example
```python
from src.litellm_client import LiteLLMClient

client = LiteLLMClient(
    base_url="http://localhost:4000",
    api_key="sk-1234567890",
    timeout=30.0,
    max_retries=2
)
```

### Response
N/A (Constructor)
```

--------------------------------

### Get MCP Client IP

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the client's IP address. Corresponds to the GET /v1/mcp/network/client-ip API endpoint.

```python
get_mcp_client_ip() -> dict
```

--------------------------------

### Health Object Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Example of a health object detailing deployment status. Includes counts of healthy and unhealthy endpoints and their details.

```json
{
    "healthy_count": 2,
    "unhealthy_count": 0,
    "healthy_endpoints": [
        {
            "name": "gpt-4o",
            "model_name": "gpt-4o",
            "litellm_params": {},
            "status": "healthy"
        }
    ],
    "unhealthy_endpoints": []
}
```

--------------------------------

### Get MCP Registry

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the MCP registry. This endpoint is optional and may return a 404 if not available. Corresponds to GET /v1/mcp/registry.json.

```python
get_mcp_registry() -> dict
```

--------------------------------

### Key Type Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Details an API key including its alias, spend, budget, and associated models. The raw token is included but should not be logged. Standard verbosity omits some full fields for brevity.

```python
{
    "token": "sk-...",
    "key_name": "key_xyz",
    "key_alias": "api-key-1",
    "spend": 25.50,
    "max_budget": 100.0,
    "models": ["gpt-4o"],
    "user_id": "user-123",
    "team_id": null,
    "expires": "2025-12-31T23:59:59Z",
    "blocked": false
}
```

--------------------------------

### GET /v1/mcp/server/{server_id}

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Retrieves details for a specific MCP server identified by its ID. Use this to get information about a particular server.

```APIDOC
## GET /v1/mcp/server/{server_id}

### Description
Get MCP server by id.

### Method
GET

### Endpoint
/v1/mcp/server/{server_id}

### Parameters
#### Path Parameters
- **server_id** (str) - Required - The unique identifier of the MCP server.

### Response
#### Success Response (200)
- **MCP server dict** - A dictionary containing the details of the specified MCP server.

#### Response Example
```json
{
  "example": "response body"
}
```
```

--------------------------------

### Embedding Object Example

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Example of an embedding object, typically returned as a list. Contains model information, embedding vectors, and usage statistics.

```json
{
    "object": "list",
    "model": "text-embedding-3-small",
    "data": [
        {
            "index": 0,
            "embedding": [0.123, -0.456, ...]
        }
    ],
    "usage": {
        "prompt_tokens": 10,
        "total_tokens": 10
    }
}
```

--------------------------------

### POST /model/new

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Registers a new model deployment with the LiteLLM proxy.

```APIDOC
## POST /model/new

### Description
Register a new model deployment.

### Method
POST

### Endpoint
/model/new

### Parameters
#### Query Parameters
None

#### Request Body
- **model_name** (str) - Required - The name of the model.
- **litellm_params** (dict) - Required - Parameters for LiteLLM.
- **model_info** (dict) - Required - Information about the model.

### Response
#### Success Response (201)
Created deployment dict

### Status Codes
201 (created), 400 (invalid), 401, 403
```

--------------------------------

### MCPSubmission Example (Standard Verbosity)

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/types.md

Represents a submission for an MCP server, including details about the server and its approval status. Use this for tracking server submissions.

```python
{
    "server_id": "server-456",
    "server_name": "Community Tool",
    "alias": "community-mcp",
    "transport": "http",
    "url": "http://example.com/mcp",
    "approval_status": "pending",
    "submitted_by": "user@example.com",
    "submitted_at": "2025-06-14T10:00:00Z"
}
```

--------------------------------

### Get MCP OAuth User Credential Status

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the status of OAuth user credentials for a specific MCP server. Corresponds to GET /v1/mcp/server/{id}/oauth-user-credential/status.

```python
get_mcp_oauth_user_credential_status(server_id: str) -> dict
```

--------------------------------

### Creating a New Project

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Creates a new project, likely within an organization.

```python
new_project_data = {"name": "Alpha Project", "organization_id": "tech-solutions"}
await client.projects.create_project(data=new_project_data)
print("Project created.")
```

--------------------------------

### Get Model Info by LiteLLM ID

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Fetches admin-side model information for all deployments or a specific one using its internal LiteLLM ID via GET /model/info. Returns the upstream payload.

```python
async def get_model_info(self, litellm_model_id: Optional[str] = None) -> dict:
    pass
```

```python
all_models = await client.get_model_info()
one_model = await client.get_model_info(litellm_model_id="123e4567")
```

--------------------------------

### Initializing LiteLLMClient

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Basic initialization of the `LiteLLMClient` with default settings. This client is used for making asynchronous HTTP requests to the LiteLLM API.

```python
from litellm import LiteLLMClient

client = LiteLLMClient()
# Use client for API calls...
```

--------------------------------

### GET /credentials/by_name/{credential_name}

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Retrieves a specific credential by its name.

```APIDOC
## GET /credentials/by_name/{credential_name}

### Description
Get a credential by name.

### Method
GET

### Endpoint
/credentials/by_name/{credential_name}

### Parameters
#### Path Parameters
- **credential_name** (str) - Required - The name of the credential to retrieve.

### Response
#### Success Response (200)
- **Credential dict** - A dictionary containing the details of the specified credential.

### Status Codes
- 200
- 401
- 403
- 404
```

--------------------------------

### Get Health Backlog

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the health backlog information for the system.

```python
check_health_backlog()
```

--------------------------------

### GET /public/model_hub

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Retrieves a list of models published to the public Model Hub.

```APIDOC
## GET /public/model_hub

### Description
List models published to the public Model Hub.

### Method
GET

### Endpoint
/public/model_hub

### Parameters
#### Query Parameters
None

#### Request Body
None

### Response
#### Success Response (200)
Upstream payload (list or dict)

### Status Codes
200, 401, 403
```

--------------------------------

### Accessing Configuration Settings

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Shows how to access and check for the presence of API key configuration settings using Pydantic's `BaseSettings`.

```python
from litellm import settings

if settings.has_api_key:
    print("API key is configured.")
else:
    print("API key is not configured.")
```

--------------------------------

### Tool Invocation with Parameters

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Shows how to invoke an MCP tool that requires specific parameters.

```python
tool_result = await client.mcp_gateway.invoke_tool(
    tool_name="get_model",
    tool_params={"model_id": "gpt-3.5-turbo"}
)
print(tool_result)
```

--------------------------------

### get_user_daily_activity

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Get daily activity for a specific internal user within a date range.

```APIDOC
## GET /user/daily/activity

### Description
Get daily activity for an internal user.

### Method
GET

### Endpoint
/user/daily/activity

### Parameters
#### Query Parameters
- **user_id** (str) - Required - The ID of the user.
- **start_date** (str) - Required - The start date for the activity log (YYYY-MM-DD).
- **end_date** (str) - Required - The end date for the activity log (YYYY-MM-DD).
```

--------------------------------

### Creating a New Organization

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Registers a new organization.

```python
new_org_data = {"name": "Tech Solutions Inc.", "owner_email": "admin@techsolutions.com"}
await client.organizations.create_organization(data=new_org_data)
print("Organization created.")
```

--------------------------------

### Get Public MCP Hub Listing

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the listing of public MCP hubs.

```python
await get_public_mcp_hub()
```

--------------------------------

### list_credentials

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Lists all provider credentials. This operation corresponds to a GET request to the /credentials endpoint.

```APIDOC
## GET /credentials

### Description
List provider credentials.

### Method
GET

### Endpoint
/credentials

### Response
#### Success Response (200)
- The upstream payload (list of credential dicts). Credential values are not included — use `get_credential()` for full details.

### Request Example
```python
# Usage:
creds = await client.list_credentials()
for cred in creds:
    print(cred["credential_name"])
```
```

--------------------------------

### get_public_hub_info

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Get Model Hub metadata, including title, description, and useful links.

```APIDOC
## GET /public/model_hub/info

### Description
Get Model Hub metadata, including title, description, and useful links.

### Method
GET

### Endpoint
/public/model_hub/info

### Response
#### Success Response (200)
- **dict** - Hub info dict.

### Response Example
```json
{
  "title": "LiteLLM Model Hub",
  "description": "A central place for discovering and managing models.",
  "useful_links": {
    "Documentation": "https://docs.litellm.ai"
  }
}
```
```

--------------------------------

### Listing Projects

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Fetches a list of projects associated with organizations or users.

```python
projects = await client.projects.list_projects()
print(projects)
```

--------------------------------

### Listing Customers

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Fetches a list of customer accounts.

```python
customers = await client.customers.list_customers()
print(customers)
```

--------------------------------

### GET /public/litellm_model_cost_map

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/endpoints.md

Retrieves LiteLLM's static map of model costs and capabilities.

```APIDOC
## GET /public/litellm_model_cost_map

### Description
Get LiteLLM's static model cost and capability map.

### Method
GET

### Endpoint
/public/litellm_model_cost_map

### Parameters
#### Query Parameters
None

#### Request Body
None

### Response
#### Success Response (200)
Large cost/capability mapping dict (~1MB)

### Status Codes
200, 401, 403
```

--------------------------------

### Generate Key with Common Fields

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/README.md

Example of generating a key using named arguments for common fields like alias, models, and max budget.

```python
# Named args for common fields
key = await generate_key(
    key_alias="api-key-1",
    models=["gpt-4o"],
    max_budget=100.0
)
```

--------------------------------

### get_budget_settings

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Retrieves the current budget settings configuration.

```APIDOC
## GET /budget/settings

### Description
Get budget settings.

### Method
GET

### Endpoint
/budget/settings
```

--------------------------------

### get_user_info

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/tools.md

Get information about an internal user. If `user_id` is omitted, returns the caller's information.

```APIDOC
## GET /user/info

### Description
Get info about an internal user. If `user_id` omitted, returns caller's info.

### Method
GET

### Endpoint
/user/info

### Parameters
#### Query Parameters
- **user_id** (Optional[str]) - Description: The ID of the user to get info for. If omitted, returns caller's info.
- **verbosity** (str) - Optional - Verbosity level ('standard' or 'detailed'). Defaults to 'standard'.
```

--------------------------------

### get_model

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Retrieves a single model entry from the OpenAI-compatible models endpoint (`GET /v1/models/{model_id}`).

```APIDOC
## get_model

### Description
Get a single model entry from the OpenAI-compatible models endpoint (`GET /v1/models/{model_id}`).

### Method
`get_model(model_id: str)`

### Endpoint
`GET /v1/models/{model_id}`

### Parameters
#### Path Parameters
- **model_id** (str) - Required - OpenAI-style model id (e.g., `gpt-4o`)

### Request Example
```python
model = await client.get_model("gpt-4o")
print(model["owned_by"])
```

### Response
#### Success Response
- **Model dict**: Contains model details.
```

--------------------------------

### Access Module Globals Settings

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/configuration.md

Accesses the singleton instance of LiteLLMSettings to retrieve configuration values like proxy URL and timeout. This instance is loaded at module import time.

```python
from src.config import settings

print(settings.proxy_url)         # http://localhost:4000
print(settings.timeout_seconds)   # 30.0
```

--------------------------------

### Add New Model Deployment

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Registers a new model deployment using POST /model/new. Requires model name, LiteLLM parameters for routing, and model information metadata.

```python
async def add_model(
    self,
    model_name: str,
    litellm_params: dict,
    model_info: dict,
) -> dict:
    pass
```

```python
deployment = await client.add_model(
    model_name="gpt-4o",
    litellm_params={
        "model": "openai/gpt-4o",
        "api_key": "sk-1234567890"
    },
    model_info={
        "id": "deployment-123",
        "db_model": False
    }
)
```

--------------------------------

### Getting Model Cost Map

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Retrieves the cost mapping for various models from the model hub.

```python
cost_map = await client.model_hub.get_model_cost_map()
print(cost_map)
```

--------------------------------

### MCP Gateway Server/Toolset Extras Argument

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/README.md

Illustrates the use of the 'extras' argument for less common fields in server and toolset operations.

```python
add_mcp_server(..., extras: dict)
update_mcp_server(..., extras: dict)
register_mcp_server(..., extras: dict)
test_mcp_connection(..., extras: dict)
```

--------------------------------

### Discovering MCP Gateway Tools

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Discovers available tools exposed by MCP gateway servers.

```python
discovered_tools = await client.mcp_gateway.discover_tools()
print(discovered_tools)
```

--------------------------------

### Tool Invocation for MCP Gateway Toolset Management

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/MANIFEST.md

Demonstrates invoking a tool for managing MCP gateway toolsets, such as creating a new toolset.

```python
toolset_create_result = await client.mcp_gateway.invoke_tool(
    tool_name="create_mcp_toolset",
    tool_params={"name": "default-tools", "tool_ids": ["list_models", "get_model"]}
)
print(toolset_create_result)
```

--------------------------------

### get_credential

Source: https://github.com/tetra-2023/litellm-mcp/blob/main/_autodocs/client.md

Retrieves a specific credential by its name. This operation corresponds to a GET request to the /credentials/by_name/{credential_name} endpoint.

```APIDOC
## GET /credentials/by_name/{credential_name}

### Description
Get a credential by name.

### Method
GET

### Endpoint
/credentials/by_name/{credential_name}

### Parameters
#### Path Parameters
- **credential_name** (str) - Required - Name of the stored credential

### Response
#### Success Response (200)
- **credential_name** (str) - Name of the stored credential.
- **credential_info** (dict) - Information about the credential.
- **credential_values** (dict) - Credential values (if verbosity='full').

### Request Example
```python
# Usage:
cred = await client.get_credential("openai-prod")
print(cred["credential_name"])
```
```