MCP lets an AI application discover and call external tools through a standard protocol. Neuroslop runs as a local stdio server: your client starts the process, reads its tool definitions, and calls only the tool you approve in that client.
neuroslop_check
Returns rule-based style signals and, when analysis applies, a 0–100 liveliness score plus concrete slop markers. The same input produces the same rule-based result; short or unsupported input may have no overall score, and the check itself does not call an LLM. Free and keyless through the hosted HTTPS API; anti-abuse rate limits apply.
neuroslop_humanize
Humanize AI drafts: reshape the delivery, make the language more natural and compare versions while checking important details. Only neuroslop_humanize requires an API key.
neuroslop_clean
Removes zero-width characters, fake spaces and mixed-script homoglyphs locally. It needs no network connection and no API key.
Install in three steps
1Check Node.js
Run node --version. Version 18 or newer is required because the server uses the built-in Fetch API.
Replace the absolute path. NEUROSLOP_API_KEY is optional and is used only by neuroslop_humanize; neuroslop_check and neuroslop_clean work without a key.
No key yet. A key is only for humanization or higher-volume HTTP API use; free MCP checks remain keyless. Get an API key
Verify the connection
Restart or reload the MCP client, then ask: “Use Neuroslop to clean hidden characters in this text, check its style signals when analysis applies, and explain the markers.” Your client should request permission to call neuroslop_clean and neuroslop_check.
Data and security boundaries
Tool
Where it runs
Data sent
Key
neuroslop_clean
Locally in Node.js
Nothing leaves the process
Not required
neuroslop_check
Official Neuroslop API
The text passed to the tool, over HTTPS
Not required · rate-limited
neuroslop_humanize
Official Neuroslop API
The text and selected rewrite options, over HTTPS
Required
The module accepts the key only from NEUROSLOP_API_KEY. It does not print the key in tool responses. Review your MCP client's own permission and logging settings before processing sensitive material.
Useful workflows
Content QA
Check a draft before publication, inspect the exact markers and applicability notice, then humanize only if the context justifies a rewrite.
Agent output guardrail
Add a final quality pass to an agent workflow without copying text between a browser, editor and chat.
Unicode hygiene
Clean pasted content before diffing, indexing or reviewing it so invisible characters cannot distort the result.
Frequently asked questions
What is the Neuroslop MCP server?
It is a local stdio process that exposes three Neuroslop text-quality tools to compatible AI agents.
Which MCP clients are supported?
Any client that can launch a standard MCP server over stdio, including Claude Code, Cursor and Codex. Exact configuration file locations are controlled by each client.
Do I need an API key?
Free and keyless through the hosted HTTPS API; anti-abuse rate limits apply. Strips hidden characters: zero-width spaces, homoglyphs. Local, no key needed. Only neuroslop_humanize requires an API key.
What data leaves my computer?
The clean tool sends nothing. Check and humanize send only the text provided to that tool and its options to the official Neuroslop API over HTTPS.
Does it require npm dependencies?
No. The server is one zero-dependency JavaScript module for Node.js 18 or newer.
Where should I store the key?
Use the MCP client's environment configuration or its supported secret store, and never commit a real key to your repository.