[grade=A] DashCaddy MCP Server — AI-native self-hosting control plane

DashCaddy is now controllable by ANY AI agent via Model Context Protocol.

17 MCP tools exposed:
- Service management: list, get, health check
- Container management: list, start/stop/restart/remove
- Deployment: deploy app, wizard recommendations, catalog search, discovery
- System: health, metrics, diagnostics
- Infrastructure: DNS listing, Caddyfile generation
- Backup & Recovery: create backup, status
- Fleet: list hosts

Protocol: JSON-RPC 2.0 over stdio
Connection: DASHCADDY_URL + DASHCADDY_API_KEY env vars

Any MCP-compatible agent (Claude Desktop, Hermes, GPT) can now:
'I want to stream movies' → wizard recommends Plex/Sonarr/Radarr
'Deploy Plex' → container + Caddyfile + DNS + health check
'Why is Plex down?' → diagnostics with structured findings
'Back up everything' → full snapshot

14 tests, 1752 total pass.

AI-NATIVE-VISION.md

@@ -0,0 +1,56 @@
+# DashCaddy AI-Native Vision
+
+## The Vision
+DashCaddy should be inherently optimized for AI agents to control it.
+Users should be able to self-host anything using natural language.
+
+## Core Principles
+1. **AI as first-class citizen** — not a bolt-on chatbot, but where the API itself is designed for AI consumption
+2. **Natural language → deployment** — "host a Plex server" → running container + reverse proxy + DNS + health check
+3. **Agent-friendly API** — structured responses, semantic error codes, state machines, idempotent operations
+4. **MCP-native** — DashCaddy should expose itself as an MCP server so any AI agent can control it
+
+## Architecture Layers
+
+### Layer 1: Natural Language Intent Router (NEW)
+`POST /api/v1/ai/intent` — Takes natural language, returns structured action plan
+- "I want to stream movies" → { category: media-streaming, recommended: [plex, sonarr, radarr] }
+- "Set up a password manager" → { category: file-sync, recommended: [vaultwarden] }
+- "Block ads on my network" → { category: home-network, recommended: [adguard] }
+- "Why is Plex down?" → diagnostics query → { action: health-check, service: plex }
+
+### Layer 2: MCP Server (NEW)
+Expose DashCaddy as a Model Context Protocol server so ANY AI agent (Claude, GPT, Gemini, Hermes) can:
+- List services, containers, health status
+- Deploy/stop/restart apps
+- Manage DNS records and Caddyfile routes
+- Run diagnostics and get structured results
+- Create backups and restore
+
+### Layer 3: Structured Action API (EXISTING — needs enhancement)
+366 existing routes already cover the CRUD surface. Enhancement needed:
+- Consistent response envelopes (already have `ok()` / `errorResponse()`)
+- All error responses include machine-readable codes (DC-086 done — 80 codes)
+- Idempotency keys for mutating operations
+- Operation receipts (UUID + status tracking)
+
+### Layer 4: Semantic Service Catalog (EXISTING — DC-104)
+76 templates with categories, auto-categorization, search.
+Enhancement: Add intent tags ("movie streaming", "password manager", "ad blocking")
+
+### Layer 5: Diagnostic Engine (NEW)
+`POST /api/v1/ai/diagnose` — Structured troubleshooting
+- "Why is X slow?" → checks: CPU, memory, network, disk I/O, container logs
+- Returns structured findings with severity + suggested fix
+- Can auto-apply fixes with user approval
+
+### Layer 6: Deployment Orchestrator (PARTIAL — DC-103 + wizard)
+"Deploy Plex" → full automation chain:
+1. Pull image
+2. Create container with optimal config
+3. Generate Caddyfile route (DC-106)
+4. Create DNS record
+5. Add to services list
+6. Start health monitoring
+7. Configure notifications
+8. Return ready-to-use URL