MCP registry submissions — paste-ready

Registries where the prompt-shield MCP server belongs. Each entry below is formatted for the registry’s expected shape and can be pasted into a PR or a form. None of these are auto-submitted — submit them yourself, in your own name, when you’re ready.

Update the version number and any counts to match the shipped release before submitting.

1. modelcontextprotocol/servers — official community list

Repo: https://github.com/modelcontextprotocol/servers How: Open a PR adding an entry to the community-servers table in the README. Category: Security / guardrails.

Paste as a new row in the community table:

- **[prompt-shield](https://github.com/mthamil107/prompt-shield)** — Prompt-injection firewall for LLM applications. Exposes `scan_input`, `scan_output`, `scan_tool_result`, and `list_detectors` tools backed by 34+ input detectors, a federated ed25519-signed threat feed, and a tool-result boundary primitive.

2. punkpeye/awesome-mcp-servers

Repo: https://github.com/punkpeye/awesome-mcp-servers How: Open a PR adding an entry alphabetically under Security (create the section if it doesn’t exist).

- [prompt-shield](https://github.com/mthamil107/prompt-shield) 🐍 🏠 - Prompt-injection firewall. Scans user input, LLM output, and tool-result content for injection, jailbreaks, PII, code injection, and toxicity via 34+ detectors with a federated ed25519-signed threat feed.

Legend used by that repo: 🐍 = Python, 🏠 = self-hosted, ☁️ = cloud, 🍎 = macOS, 🐧 = Linux, 🪟 = Windows.

3. wong2/awesome-mcp-servers

Repo: https://github.com/wong2/awesome-mcp-servers How: PR adding to Community Servers or a Security subsection.

Same content as entry #2 above works here.

4. appcypher/awesome-mcp-servers

Repo: https://github.com/appcypher/awesome-mcp-servers How: PR adding to a Security or Guardrails section.

Same content as entry #2 above.

5. Smithery.ai — hosted MCP registry

Site: https://smithery.ai/ How: Sign in with GitHub, click “Add server”, fill in the form. Smithery scrapes your repo’s smithery.yaml if present.

Suggested form values:

  • Server ID: mthamil107/prompt-shield
  • Command: prompt-shield-mcp
  • Description: Prompt-injection firewall exposing scan_input, scan_output, scan_tool_result, and list_detectors tools. Backed by 34+ detectors and a federated signed threat-intel feed.
  • Category: Security
  • License: Apache 2.0

Optionally check in a smithery.yaml at the repo root:

name: prompt-shield
description: Prompt-injection firewall for LLM applications
version: 0.8.0
runtime: python
command: prompt-shield-mcp
args: []
env:
  PROMPT_SHIELD_CONFIG:
    description: Optional path to a custom prompt-shield YAML config
    required: false

6. MCP.so registry

Site: https://mcp.so/ How: Submit via the “Submit Server” form on the site.

Same values as Smithery.

7. PulseMCP registry

Site: https://www.pulsemcp.com/ How: They index public MCP servers automatically once you’re in one of the awesome-lists above. No direct submission needed.

Suggested submission order

The goal is durable presence with minimum friction. Prioritise registries that require no ongoing maintenance:

  1. modelcontextprotocol/servers — official, most authoritative, highest discovery signal. Do this first.
  2. punkpeye/awesome-mcp-servers — largest community list on GitHub.
  3. Smithery.ai and MCP.so — hosted registries with search + install UIs. Higher upside because a Cursor user can install by clicking, not by editing JSON.
  4. The remaining awesome-lists — small individual upside; do them once and forget.

Timing note

Submit only after you have:

  • Cut a v0.8.0 release tag on GitHub with the MCP server changelog entry visible.
  • Published prompt-shield-ai==0.8.0 to PyPI so pip install "prompt-shield-ai[mcp]" picks up the console script.
  • Manually tested prompt-shield-mcp end-to-end in Claude Desktop or Cursor so the demo transcript in docs/mcp-server.md is reproducible.

Submitting before those three steps produces broken listings that maintainers close.

After the PRs merge

Add merged registry URLs to docs/mcp-server.md under a “Where to find prompt-shield” section — that’s evergreen social proof for anyone reading the docs later.