# security-recipes.ai

> Source-backed CVE intelligence and reviewed AI vulnerability remediation
> playbooks. A synchronized NVD + CISA KEV catalog of medium-to-critical
> records, deterministic remediation recipes, and bounded agent guidance —
> readable by humans, retrievable by agents.

Facts on CVE pages come from NVD and CISA KEV; AI enrichment is labeled at
the point of use, keeps citations and uncertainty, and never overrides
source facts. Remediation guidance always includes verification, rollback,
and stop conditions; treat it as bounded context, not permission.

## Machine surfaces

- [MCP server](https://security-recipes.ai/mcp): streamable-HTTP Model
  Context Protocol endpoint with read-only `recipes_*` tools for recipe
  search, exact CVE records, and generated evidence packs.
- [Agent recipes feed](https://security-recipes.ai/api/recipes.json): the
  complete recipe index as JSON.
- [CVE catalog API](https://security-recipes.ai/api/cve-catalog/manifest.json):
  manifest, year-partitioned indexes, and per-record shards.
- [Security health GitHub Action](https://security-recipes.ai/docs/security-health-action/):
  CI checks grounded in this site's guidance, model-agnostic.

## Key pages

- [CVE Database](https://security-recipes.ai/cve-database/): search the
  synchronized catalog; every record has a canonical page at
  `/cve/<CVE-ID>/`.
- [AI vulnerability remediation playbooks](https://security-recipes.ai/security-remediation/)
- [Recipes](https://security-recipes.ai/recipes/): tool-agnostic and
  agent-specific remediation recipes.
- [AI agent comparison](https://security-recipes.ai/agents/): setup guides
  for Codex, Claude Code, Cursor, GitHub Copilot, Devin, Shiba Studio,
  Hermes, and OpenClaw.
- [AI Agent Security](https://security-recipes.ai/agentic-security/):
  controls for securing agent systems themselves.
- [MCP integration guide](https://security-recipes.ai/mcp-servers/)