LiteAdmin MCP

AI agent · Workflow / back-office agent · IT & helpdesk by LiteLLM (BerriAI)

Control your LiteLLM gateway with natural language instructions.

An MCP-based AI agent that allows platform teams to control LiteLLM gateways through natural language, enabling chat-based key management, model configuration, and error troubleshooting.

Agent specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Languages
English

Category: AI Agents

Integrations
Claude Code, Codex, LiteLLM Gateway
Interface Type
Model Context Protocol (MCP)
Supported Coding Agents
Claude Code, Codex
Primary Functions
Key creation, model management, budget control, log analysis
  • Control gateway via natural language instructions
  • Create keys and manage budgets through chat
  • Troubleshoot errors by searching gateway logs
  • Built on Model Context Protocol (MCP)

Overview

LiteAdmin MCP is an AI agent interface that allows platform teams to manage their LiteLLM gateway using natural language instructions. By connecting coding agents like Claude Code or Codex to the gateway, users can perform administrative tasks such as creating virtual keys, adding new models, and controlling team budgets through a chat interface. The agent also assists with observability by searching for specific request errors within the gateway logs. This tool leverages the Model Context Protocol (MCP) to provide a conversational administrative layer for AI infrastructure. It is designed for teams that want to streamline gateway operations without manual configuration changes.

Gateway Administration

  • Key Management: The agent creates virtual keys for teams and users with specific monthly budgets and rate limits.
  • Model Configuration: Users can add new models to the gateway through natural language instructions to the agent.
  • Team Administration: The agent can be instructed to control teams and adjust budgets.

Observability and Troubleshooting

  • Error Troubleshooting: The agent can find specific requests that resulted in errors within the gateway logs.
  • Log Analysis: Administrators can ask the agent to find a request with an error.

Connectivity and Integration

  • Agent Compatibility: The tool connects Claude Code or Codex to the gateway for administrative tasks.
  • MCP Integration: The agent uses the Model Context Protocol to interface with other tools.

What it does

  • Create keys\nAdd models\nControl teams and budgets\nFind requests with errors

Use cases

  • Platform teams create virtual keys for new departments by giving a simple text command to their coding agent.
  • Infrastructure engineers troubleshoot production issues by asking the agent to find specific request errors in the gateway logs.
  • Administrators adjust monthly team budgets or add new model providers to the gateway through a conversational interface.

Specifications

Integrations
Claude Code, Codex, LiteLLM Gateway
Interface Type
Model Context Protocol (MCP)
Supported Coding Agents
Claude Code, Codex
Primary Functions
Key creation, model management, budget control, log analysis

How it works

  1. Connect to Agent: Connect Claude Code or Codex to the LiteLLM gateway using the LiteAdmin MCP interface.
  2. Issue Instructions: Provide natural language instructions to the agent, such as 'Create a key for the search team with a $200 monthly budget'.
  3. Gateway Execution: The agent executes the administrative task on the LiteLLM gateway, such as updating configurations or querying logs.

FAQ

What is LiteAdmin MCP?

LiteAdmin MCP allows you to control the LiteLLM gateway with natural language instructions to an agent, such as creating keys, adding models, or finding errors in logs.

Which agents does LiteAdmin MCP work with?

It is designed to connect with coding agents like Claude Code and Codex.

How do I manage budgets with LiteAdmin MCP?

You can tell the agent to create a key for a specific team with a set monthly budget, and it will perform the action on the gateway.

Can the agent help with troubleshooting?

Yes, you can tell the agent to find a request with an error, and it will search the gateway logs for you.

Added by AntSpace from public sources (docs.litellm.ai).

More AI agents by LiteLLM (BerriAI)