Vulnerability Intelligence Plugin

Deterministic, source-grounded CVE and CWE lookup for your agent, from CloudPeek's indexed Open Knowledge Format corpus. No account, no sign-in and no token.

Installation

Pick your client. Every route below reaches the same endpoint; installing the plugin package rather than just the MCP server is what gets you the three skills as well as the tools.

Claude Code

/plugin marketplace add Cloud-Peek/agent-plugin

/plugin install cloudpeek-vulnerability-intelligence@cloudpeek-plugins

To add only the MCP server, without the skills:

claude mcp add --transport http cloudpeek-vulnerability-intelligence https://mcp.cloudpeek.ai/vulnerability/mcp

VS Code and GitHub Copilot

Run Chat: Install Plugin From Source and give it https://github.com/Cloud-Peek/agent-plugin, or browse the Extensions view with the @agentPlugins filter.

To add only the MCP server, put this in .vscode/mcp.json:

{
  "servers": {
    "cloudpeek-vulnerability-intelligence": {
      "type": "http",
      "url": "https://mcp.cloudpeek.ai/vulnerability/mcp"
    }
  }
}

Cursor

Install the plugin from Customize in the sidebar, or add the repository as a team marketplace under Dashboard → Plugins → Import from Repo.

To add only the MCP server, put this in .cursor/mcp.json for a single project, or in ~/.cursor/mcp.json for every project:

{
  "mcpServers": {
    "cloudpeek-vulnerability-intelligence": {
      "url": "https://mcp.cloudpeek.ai/vulnerability/mcp"
    }
  }
}

Codex

Codex reads the portable package directly, so installing from the repository gives you both the server and the skills. To wire up the server on its own, add it to your Codex MCP configuration as a Streamable HTTP server pointing at https://mcp.cloudpeek.ai/vulnerability/mcp, with no authentication block.

OpenCode

OpenCode plugins are JavaScript modules rather than manifest bundles, so this installs as an MCP server. Add to opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "cloudpeek-vulnerability-intelligence": {
      "type": "remote",
      "url": "https://mcp.cloudpeek.ai/vulnerability/mcp",
      "enabled": true
    }
  }
}

Kiro, Grok Bot, OpenClaw, NanoClaw and Hermes Agent

These all load the portable package directly. Follow the setup instructions on the compatible clients list and point the client at https://github.com/Cloud-Peek/agent-plugin.

Any other MCP client

Point it at https://mcp.cloudpeek.ai/vulnerability/mcp over Streamable HTTP. No headers, no credentials, nothing else to configure. Most clients accept a generic remote-server block equivalent to:

{
  "mcpServers": {
    "cloudpeek-vulnerability-intelligence": {
      "type": "streamable-http",
      "url": "https://mcp.cloudpeek.ai/vulnerability/mcp"
    }
  }
}

If your client cannot do Streamable HTTP, it cannot connect. There is no stdio or SSE variant of this endpoint to fall back on.

One thing to know if you are wiring up a browser-based client: the endpoint restricts the Origin header, so a request carrying one is refused with 403. Desktop and CLI clients send no Origin and are unaffected.

No token, and none to leak

This endpoint is anonymous and read-only. There is no sign-in step, no account, no API key and no token to rotate or revoke, so there is nothing here to end up in a committed config file or a shared dotfile.

That is a property of what it serves rather than an oversight: the corpus is public vulnerability data compiled from vendor and MITRE advisories, identical for every caller. It holds nothing about you, so there is nothing to authorise.

Verifying the install

Ask your agent a question that needs the corpus:

What is CVE-2021-44228?

A working install returns severity, mechanism and fix, and cites the advisory each claim came from. If instead you get a general recollection with no citation, the client is answering from the model rather than the tool, which usually means the server did not connect. Ask your client to list its available tools to confirm.

Why the answers are trustworthy

Every answer traces back to the vendor or MITRE advisory it was compiled from, rather than being recalled from a model's training data. Search is full-text matching with CVE-to-CWE relationship expansion: no embeddings, no vector stores, and no model calls anywhere in the lookup path.

That matters in practice. A model asked about a CVE from memory will produce something fluent and plausible whether or not it knows the answer, and CVE identifiers are exactly the kind of near-miss detail it gets subtly wrong. A deterministic lookup either finds the record or tells you it did not, and either way you get a citation you can follow.

It also means the same question returns the same answer tomorrow, which is what makes the output usable in a report or a ticket.

Skills

Three skills ship with the plugin. They load on demand: your agent reads their descriptions, and when a request matches one it pulls in the full instructions. You do not invoke them by name, though you can ask for one explicitly.

research-vulnerability

Research a specific CVE or CWE against CloudPeek's indexed corpus and explain it with sourced detail.

Loads when you ask what a vulnerability or weakness is, how severe it is, how it works, or how to fix it. This is the everyday one.

review-recent-vulnerabilities

Brief the user on recently indexed or modified CVEs and CWEs, optionally filtered by product, vendor, severity or date.

Loads for "what is new" questions, catch-up briefings and recurring round-ups. The output is a short, scannable briefing drawn from the recently modified end of the corpus, not an undifferentiated list of everything that changed. Reach for this in a standing weekly check.

assess-product-exposure

Work out which indexed CVEs affect a given product, vendor, package or version, and turn that into a prioritised exposure summary.

Loads when you ask whether something is vulnerable, or what you should patch. It answers from indexed affected-component evidence rather than guesswork, and prioritises the result, so you get a patching order rather than a pile of CVE identifiers.

Rate limits

The endpoint is free and anonymous, so it is rate limited per client IP to keep it that way.

A well-behaved client honours Retry-After and backs off. Interactive use never comes close to 120 requests a minute; if you are hitting it, something is looping.

Limits worth knowing

It knows what has been indexed. A CVE published minutes ago may not be in the corpus yet, and the honest answer to "is this in there?" is the one the tool gives you rather than a guess.

It knows nothing about your estate. This plugin can tell you that a CVE affects a package; it cannot tell you whether you run that package, or whether anything has fired. That needs the CloudPeek Plugin, which reaches your own tenant.

Versions and updates

The package is mirrored from the CloudPeek monorepo, which is the source of truth, and its version tracks the release of the vulnerability MCP service it was published from. That service is released on its own cadence and deliberately does not ride the CloudPeek application release train, so the two plugin version lines move independently.

Because the endpoint is hosted, tool changes reach you without reinstalling anything. Refresh the marketplace only to pick up new or changed skills.