ngrok MCP server
AI model by ngrok
Connect your coding agent to your ngrok infrastructure.
The ngrok MCP server connects coding agents like Claude Code and Cursor to your ngrok infrastructure for real-time inspection and management. It enables agents to troubleshoot error codes, analyze traffic patterns, and write Traffic Policy rules autonomously using secure OAuth authentication.
Model 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: Developer Tools
- Languages
- English (for plain language prompts and agent interactions)
- Protocol
- Model Context Protocol (MCP)
- Authentication Type
- OAuth
- Connects Claude Code, Cursor, and Codex to ngrok infrastructure
- Enables agents to inspect endpoints and traffic patterns in real-time
- Allows agents to write and validate Traffic Policy rules autonomously
- Troubleshoots ERR_NGROK codes using built-in documentation access
- Uses OAuth for secure, role-based account access
Overview
The ngrok MCP server is a Model Context Protocol implementation that connects coding agents to a developer's ngrok infrastructure. It allows tools like Claude Code, Cursor, and Codex to interact directly with ngrok accounts to check usage, inspect endpoints, and investigate traffic patterns. By providing agents with real-time context about networking resources, it enables them to write Traffic Policy rules and troubleshoot ERR_NGROK codes autonomously. The server uses OAuth for secure authentication and is designed to protect sensitive data by never returning authtokens or API keys to the agent. It is built for developers who want to integrate their AI coding assistants deeply with their production and development networking environments.
Agent Capabilities
- Coding Agent Integration: The server connects coding agents like Claude Code, Codex, and Cursor to a developer's ngrok account to provide real-time infrastructure context.
- Infrastructure Inspection: Agents can check the status of endpoints, connected ngrok agents, and investigate traffic patterns in Traffic Inspector.
- Traffic Policy Management: The server allows agents to write and validate Traffic Policy rules in plain language to shape and control traffic flow.
- Automated Troubleshooting: Agents can access ngrok documentation to explain ERR_NGROK codes and troubleshoot connectivity issues autonomously.
Security and Connectivity
- OAuth Authentication: The server uses OAuth for secure sign-in, matching a user's Slack profile email to their ngrok account member role.
- Secret Protection: The server is designed to never return secret values, authtokens, or API keys to the agent to maintain security.
- Broad Client Support: The server supports clients that use OAuth client metadata documents, such as Zed, Goose, and OpenCode.
- Agent Skills Library: Developers can install agent skills like expose-localhost and secure-endpoint to give agents specific instructions for common tasks.
Use cases
- Developers ask Claude Code to investigate 500 errors and identify which endpoints are failing in real-time.
- Engineers use Cursor to generate and validate rate-limiting Traffic Policy rules using plain language instructions.
- Teams troubleshoot ERR_NGROK codes by having their coding agent read the official documentation and account status.
- DevOps engineers check if they are approaching monthly request limits by asking their agent for a usage summary.
Specifications
- Languages
- English (for plain language prompts and agent interactions)
- Protocol
- Model Context Protocol (MCP)
- Authentication Type
- OAuth
- Supported IDEs/Agents
- Claude Code, Cursor, Zed, VS Code, Gemini CLI, Amp, Cline, Codex
- Data Access Scope
- Endpoints, Traffic Inspector, Domains, TCP Addresses, Usage, Limits
How it works
- Select Agent Config: Developers pick their preferred agent (e.g., Claude Code, Cursor) to get the specific command or configuration file needed for setup.
- Add MCP Server
- Authenticate via OAuth: The agent opens a browser window for the developer to sign in to ngrok via OAuth, linking the agent to their account.
- Interact with Agent: The developer provides plain language prompts to the agent to inspect traffic, check usage, or generate networking rules.
FAQ
Which coding agents are supported?
The ngrok MCP server supports Claude Code, Codex, Cursor, Gemini CLI, Amp, Cline, Zed, Goose, and OpenCode.
How does authentication work?
The server uses OAuth to sign in; it matches your client profile email to a member of the connected ngrok account and operates with that member's role permissions.
Can the agent see my authtokens or API keys?
No, for security reasons, the server never returns secret values, authtokens, or API keys to the coding agent.
What specific data can the agent access?
The agent can look up endpoints, connected agents, traffic logs, gateway resources like domains and IP policies, and account usage limits.
Added by AntSpace from public sources (ngrok.com).