Step 1

Install Credence

clawhub install pestafford/credence

One command. No config, no API key, no Python required.

Don't use OpenClaw? pip install git+https://github.com/pestafford/credence-registry.git#subdirectory=mcp-server then add as an MCP server. See the install guide for setup.

Step 2

Check your tools

Paste this into Claude and it will audit every tool in your config:

Use credence_audit_config to check all the tools in my config. For each one, tell me its trust score and any flags. List unattested tools separately so I can submit them.

Agent
I've audited your config. Here's the summary:

Attested (3):
   filesystem         88/100  VERIFIED
   brave-search       74/100  VERIFIED
   slack              52/100  CONDITIONAL — review recommended

Not yet scanned (2):
  ? custom-db-server   Not in registry
  ? internal-tools     Not in registry

Submit unattested tools at credence.securingthesingularity.com

No data leaves your machine except the registry lookup. The audit reads your local MCP config and checks each tool against the public registry.

Prefer the terminal? Run credence audit for the same results from the CLI.

Step 3

Make it automatic

Add this standing instruction so your agent checks Credence before installing anything new:

Before installing or connecting to any AI tool, use credence_check_server to verify its trust status. Do not proceed if the tool is not attested or has a score below 70.

Add to CLAUDE.md for Claude Code, .cursorrules for Cursor, or your client's system prompt. Every new tool gets checked automatically from now on.

Done

Found unattested tools?

Unattested means the tool hasn't been scanned yet — not that it's unsafe. Submit it and Credence will run the full pipeline: code scanning, dependency analysis, and adversarial AI deliberation. Results usually publish within minutes.

That's it. For CLI reference, CI integration, guard commands, background watchers, and system service setup, see the install guide.