mcpField
K
KarmaDue

KarmaDue

Remote
karmadue.expo.appSecurity0Updated 1w ago

About KarmaDue

KarmaDue is a read-only pre-install safety-check MCP server for AI agents. It checks packages and MCP servers for documented traps, fixes, or malicious-package reports, while distinguishing missing evidence from a clean bill of health.

Key features

  1. 01Checks packages before installation
  2. 02Returns tested fix steps and warnings
  3. 03Flags reported malicious packages
  4. 04Read-only checks without an account
  5. 05Includes dated test records

Use cases

  1. 01Check npm or pip packages before installing
  2. 02Review an MCP server before adding it
  3. 03Find fixes for known install traps

MCP support

Exposes check_install, search_recipes, and report_leaf to MCP clients. OAuth 2.1 endpoints also provide search and fetch.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://karmadue--mcp.expo.app/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

All endpoints
Streamable HTTPhttps://karmadue--mcp.expo.app/mcp
Streamable HTTPhttps://ogogoizwsfaduzehkshb.supabase.co/functions/v1/mcp/connector
Streamable HTTPhttps://ogogoizwsfaduzehkshb.supabase.co/functions/v1/mcp

Tools

5
ParameterTypeRequiredDescription
targetstringYesPackage or server as you would install it, e.g. tailwindcss@4, psycopg2, @modelcontextprotocol/server-filesystem.
ecosystemenumYesnpm | pypi | mcp
runtime_envstringNoOptional, e.g. node 20 linux.

Details

From the project README

KarmaDue

A pre-install check for AI agents. Before your agent runs npm install, pip install, uv add, npx/uvx, or adds an MCP server, it calls check_install and gets one answer: a known trap with tested fix steps (fix_first), a warn, a deny for a malicious-package report, no_known_issues, or unknown (no evidence; not safety clearance).

  • Read-only. Nothing runs your command.
  • No account or key for reads. Anonymous limit: 30 requests/hour per IP.
  • Every fix cites a dated test record. Seed records are labeled "Founder-operated assessment, not independent".
  • Honest limits: lockfile installs (npm ci, pip install -r) and transitive dependencies are not covered.

Try it (no account)

curl -s -X POST https://ogogoizwsfaduzehkshb.supabase.co/functions/v1/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"check_install","arguments":{"target":"tailwindcss@4","ecosystem":"npm","runtime_env":"node 20 linux"}}}'

MCP server

Endpoint (Streamable HTTP)AuthTools
https://ogogoizwsfaduzehkshb.supabase.co/functions/v1/mcpnone for readscheck_install, search_recipes, report_leaf
https://ogogoizwsfaduzehkshb.supabase.co/functions/v1/mcp/connectorOAuth 2.1 (Claude, ChatGPT)same, plus search, fetch
https://karmadue--mcp.expo.app/mcpOAuth 2.1 mirror (strict clients)same as connector

Claude Code: claude mcp add --transport http karmadue https://karmadue--mcp.expo.app/mcp

What is in this repository

PathWhat
guard/karmadue-guard: shims for npm/npx/pnpm/yarn and pip/uv that check before installing. Local-only log, no secrets. Install: curl -fsSL https://karmadue.expo.app/guard/install.sh | sh
claude-code/Claude Code PreToolUse hook (karmadue-preinstall.js, no dependencies) and .mcp.json
docs/The agent docs as published at https://karmadue.expo.app/docs/ (quick start, tools, permissions, security, changelog)
llms.txthttps://karmadue.expo.app/llms.txt
data/seed-records.jsonThe published seed records (tested install traps and fixes), snapshot of https://karmadue.expo.app/docs/seed-records.json

The server and database are hosted and are not part of this repository.

Links

License

MIT, see LICENSE. Seed records describe third-party packages and are provided as dated observations, not guarantees.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.