- Home
- Skills
- DevOps & Cloud
- MCP Config Doctor: Config and Log Checks for MCP Clients
Works with the AI Agents you already use
MCP Config Doctor: Config and Log Checks for MCP Clients
Find known problems in MCP config files and logs: JSON errors, broken command and args, relative paths, which definition wins, and key-like values (heuristic), without printing values.
$5
MCP Config Doctor: Config and Log Checks for MCP Clients
Example session with this skill installed
My MCP servers don't show up in Claude Desktop. Check my desktop config, my project config and the MCP log, and tell me what to fix without printing any of my keys.
- Read your context and instructions
- Compiled the mcp config doctor
Summary
The agent checked the desktop config, the project config and the MCP log with the skill's scripts. It found two structure problems in the desktop config, one key-like value in a URL in the project config (reported by length only), and two known log patterns plus one error line it could not classify.
$ python scripts/mcp_doctor.py claude_desktop_config.json
FIX [source: heuristic] claude_desktop_config.json:mcpServers.notes-server.command :: command holds 3 space-separated parts; put the program in command and the rest in args (a list)
FIX [source: F5] claude_desktop_config.json:mcpServers.weather-demo.args[0] :: relative path. F5: 'Make sure the file paths included in claude_desktop_config.json are valid and that they are absolute and not relative'
note: this script reads config files only; it does not start servers and cannot tell whether a server works
MCP-DOCTOR ok=0 fix=2 warn=0 unknown=0
$ python scripts/mcp_doctor.py .mcp.json --kind project # url with a key after '#key='
FIX [source: heuristic] .mcp.json:mcpServers.s.url :: secret-looking value in url (length 53); the value is not shown. ...
MCP-DOCTOR ok=0 fix=1 warn=0 unknown=0
$ python scripts/log_triage.py mcp.log
PATTERN not-found [source: heuristic] 1 line(s): the line may mean the client could not find the program in command ...
mcp.log:2 level=error matched='ENOENT'
PATTERN disconnect [source: heuristic] 1 line(s): the connection or the server process may have ended ...
mcp.log:5 level=info matched='transport closed'
UNKNOWN 1 line(s) hold an error word but match no known pattern; open them yourself
mcp.log:6 level=error matched='error'
| Finding | Where | Fix |
|---|---|---|
| Command and arguments in one string | Desktop config, notes-server | Put the program in command and the rest in args as a list |
| Relative path in args | Desktop config, weather-demo | Use an absolute path |
| Key-like value in a URL | Project config, server s | Move the value to an environment variable and use ${VAR} |
| Program not found | Log line 2 | Check that the program in command is installed and on the app's PATH |
| Unknown error line | Log line 6 | Open that line yourself |
Next steps
- Fix the two desktop config problems, quit the app completely and start it again.
- Move the key out of the project config before you commit it, then run the doctor again until it reports 0 key FIX.
- Read the unknown log line yourself; the skill never prints log text.
Connects securely to your tools. The creator never sees your data.
About this skill
Use when an MCP server does not show up or fails to start, when you want a config file checked before you restart the app, when one server name is defined in several places, or when you want to know whether a config file holds keys in plain text.
The config check reports, by this skill's own rules (heuristic) unless a quote is given: JSON syntax errors with line and column, a missing or misspelled mcpServers key, a command written together with its arguments, args given as one string, relative paths in the desktop config, the deprecated sse transport, unset ${VAR} references, and which definition is used when a server name appears in several scopes for one project, quoting the client and MCP documentation where a rule comes from it.
Key-like values are reported by key name and length only; config values are never printed (only key names, lengths and ${VAR} names).
The log check sorts log lines into known patterns (program not found, parse error, timeout, disconnect, 401/403, port in use) and never prints log text: each finding gives the file, line number, level and matched keyword.
An optional fixed copy repairs structure only and refuses files in which the key check finds a key-like value (heuristic); use it only when the doctor reports 0 key FIX.
Do not use when you need a server tested, connectors managed in a web app, organisation-managed settings, or help building your own MCP server.
Requires: Python 3.9+ (standard library only). Reads files you point it at; starts no server and makes no network calls.
Independent: Independent tool. Not affiliated with or endorsed by Anthropic. Also not affiliated with OpenAI, Anysphere (Cursor) or Google. Product names are used only to describe compatibility.
Reads config and log files only: it does not start, test or connect to servers, so it cannot tell whether a server works. It reports these known patterns, not all MCP problems, and makes no claim about protocol versions. The key check is a heuristic and does not catch every key: unusual key forms can slip through (for example a key that starts like a path, a short key under a neutral name, a lowercase key that looks like a package name, a key with a slash in it after a path in the same string, or a URL query value made only of digits), and ordinary words can be flagged (for example a long folder name that mixes capitals, small letters and digits). --write-fixed is only safe when the doctor reports 0 key FIX; a key the check misses is copied as it is. The PATH and ${VAR} checks use the script's own environment, which can differ from the client's. Keys are never fixed automatically.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 5 days ago
Requires