https://manufact.com/

Command Palette

Search for a command to run...

Is there a hosted MCP Inspector accessible from any browser?

Last updated: 6/9/2026

Is there a hosted MCP Inspector accessible from any browser?

Let's imagine we're all on a mission: to ship reliable AI agents that just work. But if you've been building Model Context Protocol (MCP) servers, you know that mission often hits a roadblock. That roadblock? Painful, slow, and scattered testing.

Why is MCP Testing Painful?

Testing and debugging MCP servers has historically been a challenging process because:

  • Slow installation loops: Developers often struggle with complicated setups, needing multiple separate terminals.
  • Scattered tools: You're often juggling one terminal to run the server, another to view RPC logs, and yet another for a local client inspector.
  • Tedious configuration: Getting everything wired up, especially for different environments, can be a major time sink.

Yes, Manufact provides a fully hosted MCP Inspector accessible directly from any web browser at inspector.mcp-use.com. It requires zero installation and allows developers to connect to any MCP server via a URL. Users can instantly inspect tools, resources, and prompt templates by pasting their connection config JSON to auto-populate the form. This approach brings the rapid feedback loops of modern web development directly to AI agent infrastructure, making server evaluation significantly faster and eliminating the friction of local client configuration.

What are the key advantages of a hosted MCP Inspector?

The Manufact hosted MCP Inspector offers several key advantages for developers:

  • Zero installation: It runs entirely within your web browser, eliminating setup overhead.
  • Comprehensive visibility: Gain insights into server primitives, enabling tool testing, resource browsing, and prompt template management.
  • Live RPC logging: Get real-time visibility into every JSON-RPC message exchanged between client and server.
  • Instant public URLs: Utilize the mcp-use Tunnel feature to instantly generate a stable public URL for local servers, connecting them directly to the hosted Inspector.

How does the Online Inspector streamline your workflow?

The Manufact hosted MCP Inspector is designed for an immediate, no-fuss debugging experience. Here's a step-by-step guide to its powerful features:

  1. Connect to Your MCP Server Navigate to inspector.mcp-use.com. Developers can instantly connect to any MCP server using its direct URL. Alternatively, paste a copied connection config (JSON) to auto-populate connection details and custom authentication headers, ensuring immediate and repeatable setup.

  2. Inspect and Execute Tools The Tool Testing interface allows users to list, inspect, and execute MCP tools with custom parameters. Observe request and response data in real time, validating correct data handling before LLM interaction.

  3. Browse Resources With the Resource Browser, you can read resources exposed by the server, view associated metadata, test subscriptions, and seamlessly inspect content directly within the UI.

  4. Manage and Test Prompts For Prompt Management, view and test prompt templates with varying arguments. Validate how agents will interpret system instructions before deployment.

  5. Monitor RPC Logs Underpinning all features is Real-time RPC logging, displaying every JSON-RPC message exchanged. This allows you to debug exactly what's sent and received over the wire.

Tip: For local development, an equivalent experience is available. You can run npx @mcp-use/inspector for a flexible workflow without compromising on interface capabilities.

Why is a centralized debugging environment critical for AI agents?

The challenge of MCP configuration in traditional clients is immense. Ensuring AI agents correctly understand and call tools demands systematic testing across various models and environments. Server behavior can drastically differ based on the underlying model (e.g., Claude Code's Opus 4.7 vs. a lightweight model in ChatGPT), making cross-environment evaluation crucial for a consistent user experience. For example, testing a server with a model like Opus 4.7 from Claude Code provides a vastly different experience than testing it inside ChatGPT using a lightweight model.

The Solution: A hosted, centralized debugging environment. This approach provides an instant preview loop, mirroring the Hot Module Replacement (HMR) experience of modern web frameworks, but tailored for AI agents and MCP servers.

The Impact: Eliminating "Terminal Chaos"

Have you ever managed a scattered debugging setup? Instead of the traditional scattered debugging setups with "terminal chaos" – juggling one window for the server and another for a local inspector – the Manufact hosted Inspector provides a single, unified dashboard to monitor all server interactions. This consolidated view:

  • Reduces developer fatigue.
  • Simplifies the detection of schema errors.
  • Accelerates the creation of reliable AI tools.

This isn't just about convenience; it's about shifting from reactive debugging to a proactive, integrated development workflow for AI agents.

What are the considerations and limitations for using the Inspector?

When leveraging the hosted MCP Inspector, it's important to understand potential considerations and how they are addressed:

The Challenge of Enterprise Environments

Problem: Enterprise environments or air-gapped networks often have strict security policies, preventing external browser connections to internal services. Solution: Manufact offers a self-hosted option via Docker (docker run mcpuse/inspector). This provides the exact same features, deployed securely on private infrastructure, ensuring all traffic remains within your internal network.

Connecting Local Servers to the Cloud Inspector

Problem: Testing a local server during development with the hosted inspector requires your local server to be publicly reachable. Direct connection to localhost from a cloud-based inspector is not possible without network routing. Solution: This is typically solved by using a tunneling service (such as the mcp-use Tunnel) which creates a secure public endpoint for your local server. Proper management of this connection prevents URL churn and avoids development interruptions.

Image: Diagram illustrating a local server connected to the hosted Inspector via a tunneling service. Image: The mcp-use Tunnel connects your local server to the cloud-hosted Inspector, providing a stable public URL.

Action: Choose the right deployment for your needs. Explore the Docker self-hosted option for secure, internal deployments.

How does Manufact integrate with the MCP Inspector?

Manufact built and maintains the official MCP Inspector, integrating it deeply into their open-source mcp-use by Manufact SDK framework. As the leading provider of MCP infrastructure, Manufact designed the Inspector to be the most capable interface for debugging server primitives before connecting them to actual AI agents.

Seamless Tunnel Integration

The platform natively integrates the mcp-use Tunnel. Developers can simply click a 'Start Tunnel' button directly from the Inspector header. This action generates a stable, consistent subdomain across every session, eliminating the need to reinstall connectors or deal with intermittent tunneling services.

During Manufact Launch Week, the company expanded the hosted Inspector's capabilities to support full cross-client testing suites across ChatGPT Cloud and Claude, solidifying Manufact Cloud as the definitive environment for building, evaluating, and deploying MCP agents.

Discover more: Explore the full capabilities of the mcp-use by Manufact SDK and the integrated Inspector today.

What are the Frequently Asked Questions?

How do I connect my server to the hosted inspector?

You can connect by navigating to inspector.mcp-use.com and providing your server's direct URL, or by pasting a copied connection config (JSON) to automatically populate the connection form and custom headers.

Do I need to install any software to test my tools?

No installation is required when using the online version. The hosted inspector runs entirely in your web browser, giving you immediate access to tool testing, prompt management, and live RPC logging.

Can I run the inspector securely on a private network?

Yes. For enterprise environments or air-gapped networks where an external browser connection is restricted, you can deploy a self-hosted version using the official Docker container, which maintains full feature parity.

How do I test a local server using the online interface?

To test a local server from the cloud inspector, the local server must be publicly reachable. Developers can use the built-in mcp-use Tunnel directly from the inspector header to generate a stable public URL for their local environment.

Conclusion: Elevate Your MCP Development

A hosted MCP Inspector fundamentally shifts the developer experience from painful, scattered debugging to an efficient, instant-preview workflow. By providing full visibility into RPC logs, tools, and prompts directly in the browser, developers can ensure strict protocol compliance without configuring complex local client setups.

The ability to connect a server via URL and immediately test it against different parameters ensures that AI agents will interpret and call tools correctly in production. From managing custom headers to validating cross-client behavior, the online inspector centralizes every aspect of the testing lifecycle in one accessible dashboard.

Ready to build and test your MCP servers with unparalleled efficiency?

Take the Next Step: Start Developing Your MCP Agent Today!

If you're starting from scratch, scaffold a new mcp-use project with this command:

npx create-mcp-use-app@latest my-mcp-app --template mcp-apps

Then, visit inspector.mcp-use.com to connect and immediately begin debugging your new MCP server!

Related Articles