Nervous System

io.github.levelsofselfPublished by GitHub account levelsofselfv1.12.1Updated Oct 24, 2026

Governance for LLM agents: 7 enforced rules + read-only reference tools. Safe in any MCP client.

VerifiedSTDIODesktop onlyDeveloper ToolsSecurity & Monitoring

Overview

AI-generated overview

A local MCP server that lints MCP client configurations for risky patterns and returns findings with severities and a 0-100 score.

What you can do
Its main tool, audit_mcp_config, performs a deterministic offline lint of an MCP server configuration passed as an object or JSON string, checking for inline secrets in environment variables, shell wrappers, unpinned packages, auto-install flags, broad filesystem scopes, non-TLS remote transports, and unpinned Docker tags. It returns findings with severities and a 0-100 score. A second tool, check_preflight, reads a protected-files list from a config file or environment variable and fails closed if the list is unreadable. The remaining get_ tools only return reference text such as the framework, templates, and instructions.
When to use it
Use it when you want a quick, offline check of an MCP client configuration before trusting or sharing it, for example reviewing the contents of a desktop client config file. It is also useful as a preflight check against a protected-files list. It is not a general-purpose security scanner and does not enforce anything at runtime.
Requirements
Runs locally over stdio via the npm package mcp-nervous-system, for example with npx. No authentication, environment variables, or headers are declared. The preflight check optionally reads a protected-files list from a nervous-system.config.json file or the NERVOUS_SYSTEM_PROTECTED_LIST environment variable. Desktop only; no web executable.
Before you install
The package is read-only and offline, but the README describes a separate hosted, write-capable toolset and pay-per-call endpoints billed in USDC on Base; those are not part of this package and involve spending money. The audit tool reads the configuration you pass it, which may contain secrets, so review what you submit. The get_ tools return text only and do not perform the actions their older names suggested.

Installation

In SourceWeft

  1. Open Nervous System in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Nervous System

Deterministic governance lint for MCP server configurations, plus reference tooling for the 7 rules.

What each tool actually does

ToolDoes real work?
audit_mcp_configYes. Deterministic lint, zero dependencies, fully offline, never throws. Checks inline secrets in env, shell wrappers, unpinned packages, auto-install flags, broad filesystem scopes, non-TLS remote transports, unpinned docker tags. Returns findings with severities and a 0-100 score.
check_preflightYes, when configured. Reads protected_files_list from nervous-system.config.json or NERVOUS_SYSTEM_PROTECTED_LIST. With no list it says so rather than guessing. If the list is unreadable it FAILS CLOSED.
get_* (everything else)No. They return text: the framework, templates, instructions. That is why they are named get_.

Install

npx mcp-nervous-system

Free local tool: audit_mcp_config

Pass your config as config (object) or config_json (string), e.g. the contents of claude_desktop_config.json. Returns findings with severities and a 0-100 score.

Hosted API (pay-per-call, x402, USDC on Base)

The write-capable toolset runs server-side and is not in this package.

  • POST https://api.100levelup.com/x402/audit-mcp - governance lint ($0.05 USDC)
  • GET https://api.100levelup.com/x402/verify-audit - audit-chain verification ($0.005 USDC)

Discovery: /openapi.json and /llms.txt.

2.0.0 breaking changes

  • emergency_kill_switch -> get_kill_switch_instructions. It never stopped anything.
  • verify_audit_chain -> get_audit_verification_instructions. It never verified a chain.
  • dispatch_to_llm -> get_dispatch_command. It never dispatched.
  • check_preflight no longer matches hardcoded filenames from the author's own server. It reads your list.
  • Hardcoded fleet counts and violation totals removed from get_origin_story. A number baked into a package is a fossil the moment it ships.

License

MIT

Source: README.md at commit a142da0

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v1.12.1LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts