Securie MCP Server — agent-native security

roadmap

Expose Securie's scanners, CVE library, glossary, and leak playbooks to AI agents via the Model Context Protocol (MCP). Claude Code, Cursor, Continue, Zed, and any other MCP-compatible client can query Securie directly as part of its reasoning loop. The agent gains structured tools: scan a URL, look up a CVE, check a Supabase project's RLS, explain a security term — all without the user context-switching.

What it does

Gives any MCP-compatible AI agent direct access to Securie's public scanning tools, CVE library, glossary, and leak playbooks. Agents invoke `securie.scan_url('https://staging.myapp.com')` to grade a deployed app, `securie.check_rls(project_url, anon_key)` to validate a Supabase project, `securie.lookup_cve('CVE-2025-29927')` to get a structured vulnerability summary. Because MCP is agent-native (not a REST API requiring separate auth), the integration is zero-friction for agent workflows.

When to use it

Best fit: developer workflows that already include an AI agent (Claude Code, Cursor, Continue). Especially valuable for teams where the AI writes code and the human reviews — the agent can self-check against Securie before suggesting the code. Also good for developers who want to ask security questions mid-session without leaving their editor.

Limitations

Roadmap. Requires an MCP-compatible client. Currently limited to read-only operations (scans + lookups); write actions (opening fix PRs, installing the GitHub App) stay in the product UI. Per-user rate limits apply during early access.

Install

  1. Add the Securie MCP server to your agent's configuration file:
  2. { "mcpServers": { "securie": { "command": "npx", "args": ["-y", "@securie/mcp-server"] } } }
  3. Restart your agent (Claude Code, Cursor, Continue, etc.)
  4. Verify the tools are registered by asking the agent to list Securie capabilities
  5. Agent now has access to: securie.scan_url, securie.lookup_cve, securie.check_rls, securie.explain_term, securie.leak_playbook