What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
“MCP server failed” is a symptom, not a diagnosis. If you use a local server in Claude Desktop, start by checking its configuration and launch command, then fully quit and reopen Claude Desktop and inspect the MCP logs. If you use a remote connector—or Claude Code—the setup and evidence may differ, so first identify which connection is failing.
First identify which MCP connection failed
Claude Desktop supports local MCP servers and remote custom connectors through different setup paths. A local server runs as a process on your computer; a remote connector is not launched from your local Claude Desktop server configuration. The phrase “server failed” by itself does not establish which one failed or why.
- Local server or desktop extension: use the local configuration, process, permissions, restart, and log checks below.
- Remote connector: check its own connection status and the connector’s authentication or network setup. The local
claude_desktop_config.jsonchecks below may not apply. - Claude Code or another host: use that client’s configuration and diagnostic guidance. These steps are focused on Claude Desktop, not a universal fix for every MCP client.
Anthropic’s Claude Desktop local MCP guidance distinguishes desktop extensions and their policy controls from remote connectors; its troubleshooting guidance covers extension setup, connection status, and logs. For remote setup, see Anthropic’s remote connector documentation. These links are included only as the sources identified for the relevant client paths; do not assume the local-server procedure diagnoses a remote failure.
Check the local server configuration
For a manually configured local server, Claude Desktop needs a valid JSON entry under mcpServers, along with a command and arguments that actually launch that server. The Model Context Protocol’s Build an MCP Server guide recommends absolute paths. A syntactically valid file can still fail if the executable, server file, working path, or required runtime is missing or inaccessible.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsFind the configuration file
The MCP guide gives these example locations for claude_desktop_config.json:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json - Windows:
%AppData%Claudeclaude_desktop_config.json
The exact path can depend on your environment. In the file, a local server entry has the general shape below. This is a structure example, not a universal command: substitute the command and arguments required by your particular server.
{
"mcpServers": {
"my-server": {
"command": "/absolute/path/to/runtime",
"args": ["/absolute/path/to/server-file"]
}
}
}Validate syntax and paths
- Check commas, quotes, braces, and brackets; a JSON syntax error can prevent the server entry from loading.
- Confirm the server is nested inside
mcpServersand that the server name and keys are spelled correctly. - Use paths that exist on this machine, not paths copied from another operating system or account.
- On Windows, escape backslashes in JSON (for example,
C:\path\to\runtime.exe) or use forward slashes. - Check that the configured executable and server files can be read and run by the account that launches Claude Desktop.
Do not copy a random server’s command as a generic repair. Runtime, arguments, and file paths vary by server and operating system.
Run the configured command outside Claude Desktop
Use a terminal or command prompt to check whether the configured executable and arguments identify a runnable program and server file. The exact command depends on the server, runtime, and OS, so there is no safe one-line command that applies to all MCP servers.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Read the server’s own installation instructions and identify its required runtime and launch command.
- Compare that command with the configuration’s
commandandargs, including the full paths. - Run the server using those instructions outside Claude Desktop. Resolve missing-runtime, missing-file, build, or startup errors there first.
- If you maintain the server, confirm that it builds and starts without errors before troubleshooting the Claude connection.
A server that cannot start independently is unlikely to become healthy when launched by Claude. Conversely, if it starts successfully from a terminal but fails in Claude, compare the configured environment, permissions, paths, and logs.
Fully quit and restart Claude Desktop
After changing the configuration, save it and quit Claude Desktop completely before reopening it. Closing the window alone may leave the app running, so the new configuration may not take effect.
- macOS: use Cmd+Q or quit from the Claude menu.
- Windows: quit Claude Desktop from the system tray.
- Linux: quit from the tray or terminal, as applicable to your installation.
Reopen Claude Desktop and check whether the server connects and its tools are available. Anthropic also recommends restarting when extension tools do not appear. The restart is a useful configuration refresh, but it does not fix an invalid command, missing credentials, or denied permissions.
Check extension settings, credentials, permissions, and policy
If the extension is installed but its tools are unavailable, check its required configuration fields, credentials, and file paths. Anthropic recommends completing required extension fields and verifying API keys or other authentication credentials. Confirm that paths exist and are accessible to the app, rather than merely valid for a different user or terminal session.
- Authentication: check that required keys or credentials are present and current in the server’s expected configuration location.
- Filesystem access: verify the app can read the configured executable, server file, and any files the server needs.
- Operating-system permissions: look for denied-access or security prompts and check the relevant permissions.
- Managed device: ask your administrator whether desktop extensions or the relevant directory are allowed. Anthropic notes that machine-level enterprise policy can override in-app allowlist or blocklist controls.
Do not treat a policy restriction as a server bug: on a managed device, an administrator may need to change the applicable policy.
Read the logs for the failing server
When the cause is not obvious, use Claude Desktop’s Developer settings to check connection status and server logs. Anthropic recommends enabling debug logging for extension issues. The MCP guide identifies these log directories and files:
- macOS directory:
~/Library/Logs/Claude - Linux directory:
~/.config/Claude/logs/ mcp.log: general MCP connection activity and failures.mcp-server-SERVERNAME.log: stderr output from the named server; replaceSERVERNAMEwith the server’s name.
Use the first relevant error near the failed connection to guide the next check: a missing executable points back to the command or path; a permission or authentication error points to access or credentials; a server startup error points to its own runtime or implementation. Logs are diagnostic evidence, not proof of a Claude outage. The generic phrase “server failed” does not establish a current incident or a particular release bug.
Use the symptom to choose the next check
| What you see | Check next |
|---|---|
| The server does not appear in Claude Desktop | Configuration syntax, placement under mcpServers, command and absolute paths, access permissions, extension settings, then a full restart. |
| The extension appears installed but tools are missing | Fully restart Claude Desktop; then verify required fields, credentials, and paths. |
| Tools appear, but calls fail | Inspect the general MCP log and named-server log; confirm the server builds and runs successfully. |
| A server using stdio fails or behaves unpredictably | Check whether diagnostic output is being written to stdout instead of stderr. |
| A remote connector fails | Use the remote connector’s setup and status path rather than assuming the local process configuration is involved. |
These are prioritization cues, not guaranteed diagnoses. The exact error, client, and relevant log entry determine what to do next.
If you maintain a stdio MCP server, keep stdout clean
For stdio-based servers, stdout carries JSON-RPC protocol messages. Writing diagnostic text there can corrupt the protocol stream and make a server that otherwise starts appear broken. The Model Context Protocol guide states: “For STDIO-based servers: Never use println(), as it writes to standard output (stdout) by default.” Send diagnostics to stderr or a log file instead.
This is an implementation check for people who build or maintain the server; ordinary users should look at the server logs and report the relevant startup error to its maintainer.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Or skip the browser setup
ScreenshotNeo is a website screenshot API and MCP server, not a repair for Claude’s unrelated MCP server failure. If your separate task is capturing a webpage, its API can return an image or PDF with one GET request. See the ScreenshotNeo documentation.
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpimport requests
r = requests.get("https://api.screenshotneo.com/v1/shot", params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"}, timeout=90)
open("shot.webp", "wb").write(r.content)const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);ScreenshotNeo accepts cookie and consent banners and removes 60+ known consent platforms, newsletter popups, and chat widgets before capture; those steps can be turned off. Bot checks, blank pages, timeouts, failed loads, and cache hits cost nothing, with page verdict and billing details in response headers. Its MCP server provides take_screenshot, get_page_info, and capture_pdf tools for Claude, Cursor, and other MCP clients. The free plan includes 1,000 shots per month with no card; paid plans start at $5 for 3,000. See ScreenshotNeo, or sign up free for 1,000 screenshots a month with no card.
When the checklist does not isolate the failure
Keep the exact client and connection type in view. If a local server still fails after its command runs independently, configuration and permissions are verified, Claude Desktop has been fully restarted, and logs have been checked, share the relevant error with the server maintainer or client support. For a managed installation, involve the administrator if policy may restrict extensions. For a remote connector or Claude Code, use support and documentation for that specific path rather than applying local Desktop assumptions.
Best Value
Frequently Asked Questions
Does “MCP server failed” mean Claude is down?
No. The message alone does not establish a service incident; check the affected server’s status and logs.
Should I use the local Desktop config steps for a remote connector?
No. Remote connectors use a different setup path; consult the connector’s own configuration and status guidance.
Why can I run the server manually but Claude Desktop cannot connect?
Compare the configured command, paths, credentials, and permissions with the environment used to launch Claude Desktop, then inspect its MCP logs.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →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.
