Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Connecting a Model Context Protocol (MCP) server is a two-part job: the server must be reachable, and the AI client must be configured to discover and authorize its tools. The correct steps depend on the client (an API integration, ChatGPT custom app, VS Code, or Claude), whether the server is public or private, and which credentials and actions it requires.
This guide covers those decisions using documentation current to October 2026, then gives a client-specific setup path and a troubleshooting sequence that starts with server status and logs.
What an MCP connection actually does
MCP is a protocol that lets an AI application use tools and data supplied by an external server. A server might expose search, database, file, ticketing, or deployment operations. Installing or running that server does not automatically connect it to an assistant: the selected client still needs the server endpoint, transport details, authentication, trust approval, and an enabled tool list.
Keep these layers separate:
- Server: the process or hosted service that implements MCP and declares its tools.
- Network path: a public HTTPS endpoint, a private tunnel, or a local process started by a desktop application.
- Client: the API, ChatGPT, VS Code, Claude, or another MCP-compatible application.
- Identity: OAuth, an API key, or another credential the server uses to access the connected service.
- Client controls: tool enablement, workspace policy, trust prompts, and approval requests for individual calls.
A successful connection means the client can reach the server, complete authentication, read its tool definitions, and present eligible tools to the model. It does not mean every tool is safe, enabled, or authorized to perform every action.
#1 Best Overall
- Includes Raspberry Pi 5 with 2.4Ghz 64-bit quad-core CPU (8GB RAM)
- Includes 128GB Micro SD Card pre-loaded with 64-bit Raspberry Pi OS, USB MicroSD Card Reader
- CanaKit Turbine Black Case for the Raspberry Pi 5
- CanaKit Low Noise Bearing System Fan
- Mega Heat Sink - Black Anodized
Make three decisions before configuring anything
1. Which assistant or client are you configuring?
Setup instructions are not interchangeable. OpenAI’s Responses API uses an MCP tool in an API request; a ChatGPT custom app has account and workspace controls; VS Code uses its MCP gallery and configuration files; Claude uses a remote custom-connector flow. A configuration copied from one client may use the wrong transport, field names, authentication method, or permission model in another.
2. Is the server public, private, or local?
- Public remote: the client can reach a stable URL on the public internet. You must secure the endpoint and authenticate requests as appropriate.
- Private or on-premises: the server remains behind a firewall or on an internal network. A supported secure tunnel can provide access without publishing the server directly.
- Local process: a desktop client starts a command on your machine. This offers local network access but means the process can execute with the permissions of your operating-system account.
OpenAI documents Secure MCP Tunnel for private, on-premises, or firewalled servers that should not be publicly exposed: OpenAI Secure MCP Tunnel. A tunnel changes the network path, not the server’s authorization rules.
3. Which credentials and actions are required?
Some servers are anonymous and read-only; others require OAuth or another credential and expose create, modify, or delete operations. Decide whether users should authenticate individually or whether a shared service identity is acceptable. Then decide which tools should be enabled and whether every potentially consequential call requires approval.
Preparation checklist
Collect the following before opening the client UI or writing an API request:
- The server’s exact public URL, tunnel identifier, or local start command.
- The MCP transport and protocol details required by that server and client.
- OAuth authorization information, API key, or other credential, if required.
- A list of tools the assistant actually needs, including whether each reads, creates, changes, or deletes data.
- Firewall, proxy, DNS, and allowlist requirements. Cloud-hosted clients must be able to reach the server from their own infrastructure, not merely from your laptop.
- The account plan, workspace role, and administrator policy that govern custom apps or connectors.
- A rollback plan: how to disable the app, revoke the token, stop the local process, or remove the server configuration.
Use a test account or least-privileged credential first. Do not paste a long-lived secret into a prompt or commit it to a configuration file.
Connect an MCP server with the OpenAI Responses API
The Responses API represents MCP as a tool configuration. The official guide is at MCP servers | OpenAI API.
Rank #2
- Fully assembled for plug-and-play operation
- Includes Raspberry Pi 5 with 8GB RAM
- 256 GB PCIe Pi NVMe SSD (Pre-loaded with Pi 64-Bit OS)
- M.2 HAT+
- CanaKit Turbine Black Case for the Pi 5
For a public remote server
- Create a Responses API request containing an MCP tool entry.
- Give the tool a label so it is identifiable in logs and model context.
- Set the server’s remote URL in the
server_urlfield. Use the exact route supplied by the server operator; do not substitute a homepage URL. - If the server requires OAuth, provide the access token through the request’s
authorizationvalue using the method documented by that server. - Configure approval behavior so calls are either reviewed by a user or automatically allowed only when the risk is understood.
- Send a test request that exercises a harmless read operation, then inspect the returned tool activity and your server logs.
The public OpenAI Docs MCP example is read-only and does not require authentication. That example is not evidence that other MCP servers are anonymous: follow each server’s authentication requirements.
For a private server through Secure MCP Tunnel
- Set up the Secure MCP Tunnel according to the current OpenAI instructions and keep the server on its private network.
- Use the tunnel identifier in the MCP tool configuration’s
tunnel_idrather than publishing a public URL. - Configure the server’s own authentication, including OAuth if required.
- Confirm that the tunnel process, firewall rules, and server process are running before sending the API request.
- Run a read-only test and compare the API response with tunnel and server logs.
A private tunnel reduces direct exposure but does not remove the need to vet the server, scope credentials, and approve risky tools. OpenAI’s documentation also explains approvals as a way to control and see data sent to a remote MCP server.
Add an MCP server as a ChatGPT custom app
ChatGPT custom MCP apps have a separate product flow from the Responses API. The current account and workspace guidance is in Developer mode and MCP apps in ChatGPT. Labels and eligibility can change, so check that page and your administrator’s policy immediately before setup.
- Open ChatGPT on the web and enter the developer or custom-app area available to your account.
- Choose the option to add an MCP app.
- Enter the server URL, or select the available secure tunnel when the server is private.
- Configure the authentication method requested by the server. If OAuth is used, complete the authorization flow and inspect the scopes before accepting.
- Read the risk and data-flow information shown by ChatGPT. Confirm that the app’s declared tools match the work you intend to perform.
- Create the app, then verify which tools and actions are enabled in the workspace.
- Test with a non-destructive request and confirm that approval prompts appear for actions that can change data.
Access is controlled by more than the URL. Account plan, developer-mode availability, workspace role, administrator restrictions, and app security settings all matter. At the time of the October 2026 documentation, OpenAI describes full MCP write support as rolling out in beta for Business and Enterprise/Edu, while Pro users have read/fetch permissions in developer mode. Treat those as documented rollout conditions, not a universal entitlement.
Workspace administrators can restrict app access and actions. If you cannot see the developer or app controls, ask an administrator to verify the plan and policy rather than repeatedly changing the server.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsAdd an MCP server in Visual Studio Code
VS Code’s current setup and trust guidance is in Add and manage MCP servers in VS Code.
Rank #3
- Pi5 8GB Pack: RasTech Pi 5 8GB kit includes 1 x Pi5 8GB board ,1 x 64GB Card, 2 x Card Readers,1 x Active Cooler,1 x Case for Pi5, 2 x 4K Micro HD Out Cable,1 x GaN 27W 5A USB-C Power supply,1 x Screwdriver and 1 x instructions.
- Pi5 8GB Board: The Pi5 board is equipped with a 64-bit quad-core Arm Cortex-A76 processor running at 2.4GHz and an 800MHz VideoCore VII GPU with support for OpenGL ES 3.1 and Vulkan 1.2, which delivers a significant increase in graphics performance. Dual HD Out 4Kp60 display outputs and a built-in dual 4-channel MIPI camera/display transceiver provide state-of-the-art camera support. The Pi 5 offers a 2-3 times increase in CPU performance compare to Pi4.
- Important Graphics Features: Equipped with an 800MHz VideoCore VII GPU and providing better graphics performance, suitable for multimedia applications,gaming,and graphics intensive tasks.Provides 1 UART interface,1 card slot that supports high-speed operation, 2 USB. 3 0.5 ports that support synchronous 0Gbps operation,2 USB 2.0 port ports,2 4Kp60 display outputs that support HDR.Built-in dedicated dual 4-channel 1Gbps MIPI DSI/CSI connectors,triple the total bandwidth.
- Cooling Kit for Pi 5: Compatible with Active Cooler for Raspberry Pi5, It can provide Pi 5 board with better cooling effect in using. The Case can accurately access usb-c power jack,Micro HD Out ports, usb ports, Ethernet jack, card slot, power button, 4-lane MIPI DSI/CSI connectors and so on, and it also supports installation of cooling fan.
- 64GB Card Kit and GaN 27W USB-C Power Supply: With extra 64GB card to store more files and card readers for multiple medium, keep better performance for Raspberry Pi 5, 27W USB C Power Supply is Compatible with Pi5 8GB, offers a variety of output voltage options, including 5.1V at 5A, 9.0V at 3.0A, 12.0V at 2.25A, and 15.0V at 1.8A, providing for different device requirements.
Install from the MCP gallery
- Open the Extensions view and locate the MCP server in the MCP or Extensions gallery.
- Review the publisher, declared configuration, required credentials, and requested tools.
- Install the server and accept only the trust prompts you understand.
- Open Chat and use Configure Tools to inspect the tools exposed by the server.
- Enable only the tools required for the current task.
- Run a harmless read operation and verify the result before attempting a write.
Configure a server manually
- Open the MCP configuration file used by your VS Code installation or workspace.
- Add the server definition supplied by its publisher, preserving the command, arguments, environment variables, and transport exactly.
- Keep secrets outside source control and use the credential mechanism recommended for your operating system or organization.
- Save the file, return to Chat, and use Configure Tools to confirm discovery.
- When prompted, review whether the server is trusted before allowing it to start.
Workspace trust can affect servers configured at the workspace level. A local MCP server can execute code on your machine, read files available to its process, or use network access granted to it. Inspect the publisher and every command or argument before starting it; an approval dialog is not a security audit.
Connect Claude to a remote MCP server
Claude’s supported route is a remote custom connector. Follow the current client or workspace instructions in Get started with custom connectors using remote MCP.
- Open the custom-connector area available to your Claude account or workspace.
- Enter the remote MCP server endpoint and start the authentication flow.
- Read every requested OAuth scope. Grant only the access needed for the connector’s declared tools.
- Review tool-approval prompts and keep repeated approval enabled for tools that can change or delete data unless there is a documented reason not to.
- Test a read-only operation, then inspect the server’s logs and the connector’s response.
Claude connects to a remote MCP server from Anthropic’s cloud infrastructure. Your server therefore needs to allow the relevant cloud-origin traffic, and the data path may not be the same as a test made from your local computer. Plan firewall allowlisting and data residency accordingly.
Identity design is important: a shared API key can give every connector user the shared credential’s access, whereas individually delegated OAuth access can enforce per-user permissions. Choose deliberately and document who can revoke or rotate the credential.
Understand the two permission boundaries
Server-side authorization
The MCP server may use OAuth, an API key, or another credential to reach an external service. That credential determines what the server can read or change. Scope it to the smallest practical set of resources and operations, set an expiration or rotation process, and revoke it when the connector is retired.
Assistant-side tool control
The client decides which declared tools are visible to the model and whether a human must approve a call. A server can expose read, create, modify, and delete functions; calling it “read-only” is accurate only when its implementation and enabled tool list truly prohibit writes. Recheck the tool list after a server update because capabilities can change.
Rank #4
- [ULTIMATE RASPBERRY PI 5 CASE & MINI PC] - Unlock the full potential of your Raspberry Pi 5 with the Pironman 5-MAX — the most advanced Raspberry Pi 5 Case for power users. This high-performance Raspberry Pi 5 Cooling Case features dual NVMe M.2 slots with RAID 0/1 support, AI accelerator compatibility ( e.g. Hailo-8l M.2 AI), a PCIe Gen2 switch, a PWM tower cooler + dual RGB fans and a smart OLED display. With its dual transparent panels and optimized cable management (including full-size HDMI), it’s the ideal Raspberry Pi 5 Enclosure for building a high-speed NAS, AI edge computing device, or Home Assistant hub. (Raspberry Pi NOT Included)
- [DUAL NVMe M.2 SLITS & NAS RAID SUPPORT] - Supercharge your storage with the best Raspberry Pi 5 NVMe Case solution. Featuring two expandable NVMe M.2 slots (2230-2280) powered by a built-in PCIe Gen2 switch, this Raspberry Pi 5 NAS Case supports RAID 0/1 for ultra-fast data setups. Whether you're using a high-speed NVMe SSD or a Hailo-8L AI accelerator, Pironman 5-MAX delivers the ultimate performance boost for advanced Raspberry Pi 5 AI applications and edge computing
- [ADVANCED COOLING SYSTEM] - Engineered for high-performance builds, Pironman 5-MAX features a powerful tower cooler, one PWM fan, and dual RGB fans for enhanced airflow. The dual transparent panel design improves ventilation while showcasing vibrant RGB lighting. Ideal for cooling both the Raspberry Pi 5 and dual NVMe SSDs or AI accelerators like Hailo-8L, it ensures stable operation under heavy workloads with low noise and long-term durability
- [SMART OLED DISPLAY WITH VIBRATION WAKE-UP] - Pironman 5-MAX features a 0.96" OLED screen that delivers real-time system insights including CPU usage, memory, temperature, IP address, and disk status. With customizable display options and auto sleep mode, the screen can be instantly reactivated by a light tap thanks to the built-in vibration sensor—offering a smarter and more interactive experience
- [ENHANCED FUNCTIONALITY] - Pironman 5-MAX empowers your Raspberry Pi 5 with advanced features like safe shutdown via a metal power button, customizable RGB lighting, dual full-size HDMI ports, vibration-triggered OLED wake-up, and an external GPIO extender. It also includes RTC battery support for timekeeping and seamless Home Assistant integration. With detailed guides, online tutorials, and full technical support from SunFounder, setup and use are effortless and worry-free
Approval prompts improve visibility and control, but they do not make an untrusted server safe. OpenAI and Anthropic both advise connecting only to servers you trust and evaluating their safety. A prompt can show that an action is about to occur; it cannot prove that the server will handle data correctly.
Minimum-permission review
- Disable tools unrelated to the task.
- Prefer a read-only role for discovery and testing.
- Require approval for external messages, purchases, deployments, record changes, and deletion.
- Use separate credentials for development and production.
- Record which user, workspace, or service identity authorized each connection.
- Rotate or revoke credentials when a user leaves, a server changes ownership, or a tool is removed.
Public URL, private tunnel, or local process?
The best deployment is a security and operations decision, not merely a setup convenience.
| Option | Network reachability | Exposure and data flow | Identity and control questions | Typical fit |
|---|---|---|---|---|
| Public remote URL | Reachable over the public internet from the client | Server is internet-facing; protect TLS, routing, rate limits, and authentication | Who operates the endpoint, rotates credentials, and approves tool changes? | Hosted services and APIs intended for multiple clients |
| Private secure tunnel | Client reaches a tunnel while the server stays on a private network | Direct public exposure is avoided; traffic still crosses the tunnel and reaches the private server | Who controls tunnel credentials, firewall allowlists, and the private host? | On-premises, firewalled, or internal systems |
| Local process | The desktop client launches the server on the user’s machine | Process can access local files, credentials, and network resources granted by the operating system | Can the publisher and command be trusted? Is workspace trust enabled? | Developer workflows and local repositories |
Client support differs across these options. The Responses API documents both a public server_url and a private tunnel_id; VS Code commonly starts local processes or uses configured endpoints; ChatGPT and Claude expose their own account, workspace, and remote-connector controls. Do not assume that a tunnel or local command accepted by one client is supported by another.
Troubleshoot an MCP server that will not connect
Start with evidence from the server and the client, not with repeated reinstallations. First establish whether the process is running, then whether the client can reach it, then whether authentication, tool discovery, and invocation succeed.
VS Code’s documented sequence
- Open the Command Palette and run MCP: List Servers.
- Select the affected server and inspect its status.
- Choose Show Output and read the startup, transport, authentication, and tool-discovery messages.
- Correct the exact configuration or connection error reported in the output.
- Select Restart Server.
- Retry the chat request and confirm that the expected tools are listed under Configure Tools.
These steps are documented in Troubleshoot AI in Visual Studio Code. If the process is running but no tool is invoked, treat that as a discovery or invocation problem rather than only a process-start problem.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
General diagnostic checklist for other clients
- Server status: verify the process, health endpoint if one exists, and recent server logs.
- Endpoint: check the exact URL path, tunnel identifier, port, TLS certificate, and transport expected by the client.
- Network origin: test reachability from the client-side infrastructure. A URL that works in a browser on your laptop may fail for a cloud-origin connector.
- Authentication: confirm the token is current, the OAuth redirect completed, scopes are sufficient, and the credential was not accidentally copied with whitespace or an expired value.
- Tool list: verify that the server returned tool definitions and that the client has not disabled them by policy or user setting.
- Approval state: look for a pending consent request, blocked write action, or administrator policy.
- Logs together: compare the client error timestamp with tunnel, proxy, and server logs. A client timeout with no server request usually indicates routing or firewall trouble; a server-side authorization error indicates the request arrived.
Common symptoms and targeted fixes
| Symptom | Likely layer | What to check |
|---|---|---|
| Connection refused or timeout | Process or network | Is the server listening, is the port correct, and can the client origin reach the route through the firewall or tunnel? |
| 401 or 403 response | Authentication or authorization | Token validity, OAuth scopes, audience, server-side role, and whether a shared credential has the intended access |
| Server connects but no tools appear | Protocol or discovery | Transport compatibility, server initialization errors, client tool enablement, and the server’s returned tool definitions |
| Tools appear but calls are blocked | Approval or policy | Pending user approval, workspace restrictions, disabled tools, or a write action not allowed by the account plan |
| Tool runs locally but not in a cloud client | Network origin or data policy | Cloud allowlisting, private DNS, outbound firewall rules, proxy requirements, and region or residency restrictions |
| VS Code starts a container and then loses it | Process arguments or lifecycle | Check command-line arguments and ensure the Docker container is not running detached when VS Code expects to manage its process |
When a tool is discovered but not invoked
Discovery proves only that the client received a declaration. Check that the tool is enabled, its input schema is valid, the user request actually matches its description, and an approval prompt was not dismissed or hidden by policy. Review the model request and the server’s invocation log. If the tool’s schema or name changed after an update, refresh or restart the client and compare the newly declared definition with the server documentation.
Operate the connection safely after setup
- Pin or review server versions where your client supports it, and test updates in a non-production workspace.
- Recheck the declared tools and OAuth scopes after every server or connector update.
- Monitor failed authentication, unusual tool calls, and unexpected data volume.
- Keep local server commands and environment variables out of shared repositories unless they contain no secrets.
- Document the owner, endpoint, credential type, enabled tools, approval policy, and revocation procedure.
- Remove unused connectors promptly and revoke their credentials.
Product support, rollout status, plan access, interface labels, and configuration fields can change. Before publishing or deploying a connection, verify the current documentation for the selected client and the policies of the account or workspace in which it will run.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
