docs: add Docker Compose setup for OAuth multi-user mode

- Update Dockerfile to default to streamable-http transport
- Add docker-compose.oauth-standalone.yml for existing OIDC providers
- Add .env.oauth.example template with all required settings
- Update README_DOCKER.md with OAuth deployment instructions

OAuth mode requires streamable-http transport (not stdio) since
authentication happens via browser redirect flow.

.env.oauth.example

@@ -0,0 +1,82 @@
+# mcvsphere OAuth Configuration Template
+# Copy to .env and configure for your environment
+#
+# Usage:
+#   cp .env.oauth.example .env
+#   # Edit .env with your values
+#   docker compose -f docker-compose.oauth-standalone.yml up -d
+
+# ─────────────────────────────────────────────────────────────────────────────
+# Docker Compose
+# ─────────────────────────────────────────────────────────────────────────────
+COMPOSE_PROJECT_NAME=mcvsphere
+
+# ─────────────────────────────────────────────────────────────────────────────
+# vCenter Connection (Required)
+# ─────────────────────────────────────────────────────────────────────────────
+VCENTER_HOST=vcenter.example.com
+VCENTER_USER=mcpservice@vsphere.local
+VCENTER_PASSWORD=your-secure-password
+
+# Optional: Skip SSL verification (dev only - use false in production)
+VCENTER_INSECURE=false
+
+# Optional: Specify defaults (auto-detected if not set)
+# VCENTER_DATACENTER=Datacenter
+# VCENTER_CLUSTER=Cluster
+# VCENTER_DATASTORE=datastore1
+# VCENTER_NETWORK=VM Network
+
+# ─────────────────────────────────────────────────────────────────────────────
+# MCP Transport (Required for OAuth)
+# ─────────────────────────────────────────────────────────────────────────────
+MCP_TRANSPORT=streamable-http
+MCP_HOST=0.0.0.0
+MCP_PORT=8080
+
+# Your public domain (must match Caddy proxy)
+MCP_DOMAIN=mcp.example.com
+
+# TLS mode: 'internal' for self-signed (dev), remove for auto HTTPS (prod)
+MCP_TLS_MODE=internal
+
+# ─────────────────────────────────────────────────────────────────────────────
+# OAuth / OIDC Provider (Required)
+# ─────────────────────────────────────────────────────────────────────────────
+OAUTH_ENABLED=true
+
+# OIDC Discovery URL (ends with /.well-known/openid-configuration)
+# Examples:
+#   Authentik: https://auth.example.com/application/o/mcvsphere/
+#   Keycloak:  https://keycloak.example.com/realms/myrealm
+#   Auth0:     https://myapp.auth0.com/
+#   Okta:      https://myorg.okta.com/oauth2/default
+OAUTH_ISSUER_URL=https://auth.example.com/application/o/mcvsphere/
+
+# OAuth Client Credentials (from your OIDC provider)
+OAUTH_CLIENT_ID=your-client-id
+OAUTH_CLIENT_SECRET=your-client-secret
+
+# Public callback URL (must be accessible from browser)
+OAUTH_BASE_URL=https://mcp.example.com
+
+# ─────────────────────────────────────────────────────────────────────────────
+# RBAC Permission Groups
+# ─────────────────────────────────────────────────────────────────────────────
+# Create these groups in your OIDC provider and assign users:
+#
+# | Group                  | Access Level                    |
+# |------------------------|---------------------------------|
+# | vsphere-super-admins   | Full control (all 94 tools)     |
+# | vsphere-host-admins    | Host operations + VM management |
+# | vsphere-admins         | VM lifecycle management         |
+# | vsphere-operators      | Power ops + snapshots           |
+# | vsphere-readers        | Read-only                       |
+#
+# Users without any vsphere-* group will be denied access (default-deny).
+
+# ─────────────────────────────────────────────────────────────────────────────
+# Optional Settings
+# ─────────────────────────────────────────────────────────────────────────────
+# LOG_LEVEL=INFO
+# MCVSPHERE_VERSION=0.2.2

Dockerfile

@@ -48,8 +48,11 @@ ENV PATH="/app/.venv/bin:$PATH"
 ENV PYTHONUNBUFFERED=1
 ENV PYTHONDONTWRITEBYTECODE=1
 
-# Default to SSE transport for Docker
+# Default transport - override with MCP_TRANSPORT env var
-ENV MCP_TRANSPORT=sse
+# stdio: Direct CLI usage (default for local)
+# sse: Server-Sent Events (legacy HTTP)
+# streamable-http: HTTP with OAuth support (required for multi-user)
+ENV MCP_TRANSPORT=streamable-http
 ENV MCP_HOST=0.0.0.0
 ENV MCP_PORT=8080
 
@@ -58,10 +61,9 @@ USER mcpuser
 
 EXPOSE 8080
 
-# Health check
+# Health check - works with both SSE and streamable-http
 HEALTHCHECK --interval=30s --timeout=10s --start-period=10s --retries=3 \
-    CMD python -c "import urllib.request; urllib.request.urlopen('http://localhost:8080')" || exit 1
+    CMD python -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8080/.well-known/oauth-authorization-server')" || exit 1
 
-# Run the MCP server
+# Run the MCP server - transport configured via environment
 ENTRYPOINT ["mcvsphere"]
-CMD ["--transport", "sse"]

OAUTH-ARCHITECTURE.md

@@ -1,17 +1,17 @@
-# OAuth Architecture for vSphere MCP Server
+# OAuth & RBAC Architecture for mcvsphere
 
-## The Problem
+## Overview
 
-We need to add authentication to the MCP server so that:
+mcvsphere supports multi-user OAuth 2.1 authentication with Role-Based Access Control (RBAC). This enables:
-1. Users authenticate via **OAuth 2.1 / OIDC** (using Authentik as IdP)
-2. The MCP server knows WHO is making requests (for audit logging)
-3. vCenter permissions are respected per-user
 
-**Challenge:** vCenter 7.0.3 doesn't support OAuth token exchange (RFC 8693), so we can't pass OAuth tokens directly to vCenter.
+1. **Single Sign-On** via any OIDC provider (Authentik, Keycloak, Auth0, etc.)
+2. **User Identity** for audit logging - know WHO made each request
+3. **Group-Based Permissions** - control what users can do based on OAuth groups
+4. **Audit Trail** - every tool invocation logged with user identity and timing
 
 ---
 
-## Architecture Overview
+## Architecture
 
 ```
 ┌─────────────────────────────────────────────────────────────────┐
@@ -22,360 +22,352 @@ We need to add authentication to the MCP server so that:
                              │    (browser opens for login)
                              │
 ┌────────────────────────────▼────────────────────────────────────┐
-│                         Authentik                                │
+│                      OIDC Provider                               │
-│                    (Self-hosted OIDC IdP)                        │
+│              (Authentik, Keycloak, Auth0, etc.)                  │
 │                                                                  │
 │   - Issues JWT access tokens                                     │
 │   - Validates user credentials                                   │
-│   - Includes user identity in token (sub, email, groups)         │
+│   - Includes groups claim in token                               │
 └────────────────────────────┬────────────────────────────────────┘
                              │
                              │ 2. JWT Bearer token
                              │    Authorization: Bearer <jwt>
                              │
 ┌────────────────────────────▼────────────────────────────────────┐
-│                     vSphere MCP Server                           │
+│                        mcvsphere                                 │
 │                   (FastMCP + pyvmomi)                            │
 │                                                                  │
 │  ┌─────────────────────────────────────────────────────────┐    │
 │  │                   OIDCProxy (FastMCP)                    │    │
-│  │  - Validates JWT signature via Authentik JWKS            │    │
+│  │  - Validates JWT signature via JWKS endpoint             │    │
-│  │  - Extracts user identity (preferred_username)           │    │
+│  │  - Extracts user identity (preferred_username, email)    │    │
-│  │  - Makes user available via ctx.request_context.user     │    │
+│  │  - Extracts groups from token claims                     │    │
 │  └─────────────────────────────────────────────────────────┘    │
 │                              │                                   │
 │                              ▼                                   │
 │  ┌─────────────────────────────────────────────────────────┐    │
-│  │                 Credential Broker                        │    │
+│  │                   RBACMiddleware                         │    │
-│  │  - Maps OAuth user → vCenter credentials                 │    │
+│  │  - Intercepts ALL tool calls via on_call_tool()          │    │
-│  │  - Caches pyvmomi connections per-user                   │    │
+│  │  - Maps OAuth groups → Permission levels                 │    │
-│  │  - Retrieves passwords from Vault / env vars             │    │
+│  │  - Denies access if user lacks required permission       │    │
+│  │  - Logs audit events with user identity                  │    │
 │  └─────────────────────────────────────────────────────────┘    │
 │                              │                                   │
 │                              ▼                                   │
 │  ┌─────────────────────────────────────────────────────────┐    │
-│  │                    Audit Logger                          │    │
+│  │                   VMware Tools (94)                      │    │
-│  │  - Logs all tool invocations with OAuth identity         │    │
+│  │  - Execute vCenter/ESXi operations via pyvmomi           │    │
-│  │  - "User ryan@example.com powered on VM web-server"      │    │
+│  │  - Single service account connection to vCenter          │    │
 │  └─────────────────────────────────────────────────────────┘    │
 └────────────────────────────┬────────────────────────────────────┘
                              │
-                             │ 3. pyvmomi (as mapped user)
+                             │ 3. pyvmomi (service account)
                              │
 ┌────────────────────────────▼────────────────────────────────────┐
-│                        vCenter 7.0.3                             │
+│                     vCenter / ESXi                               │
-│  - Receives API calls as the actual user                        │
+│  - Receives API calls as service account                        │
-│  - Native audit logs show real user identity                    │
+│  - mcvsphere audit logs show real user identity                 │
-│  - vCenter permissions apply naturally                          │
 └─────────────────────────────────────────────────────────────────┘
 ```
 
 ---
 
-## User Mapping Strategies
+## RBAC Permission Model
 
-Since we can't exchange OAuth tokens for vCenter tokens, we need a "credential broker":
+### Permission Levels
 
-| Strategy | How it Works | Security | Use Case |
+mcvsphere defines 5 permission levels, from least to most privileged:
-|----------|--------------|----------|----------|
-| **Service Account** | All requests use one vCenter admin account | Medium | Simple/dev |
-| **Per-User Mapping** | Map OAuth username → vCenter credentials from Vault | High | Production |
-| **LDAP Sync** | Same username/password in Authentik and vCenter SSO | Medium | AD environments |
 
-### Recommended: Per-User Mapping with Fallback
+| Level | Description | Example Tools |
+|-------|-------------|---------------|
+| `READ_ONLY` | View-only operations | `list_vms`, `get_vm_info`, `vm_screenshot` |
+| `POWER_OPS` | Power and snapshot operations | `power_on`, `create_snapshot`, `reboot_guest` |
+| `VM_LIFECYCLE` | Create/delete/modify VMs | `create_vm`, `clone_vm`, `add_disk`, `deploy_ovf` |
+| `HOST_ADMIN` | ESXi host operations | `reboot_host`, `enter_maintenance_mode` |
+| `FULL_ADMIN` | Everything including guest OS ops | `run_command_in_guest`, `restart_service` |
+
+### OAuth Groups → Permissions
+
+Users are granted permissions based on their OAuth group memberships:
+
+| OAuth Group | Permissions Granted |
+|-------------|---------------------|
+| `vsphere-readers` | READ_ONLY |
+| `vsphere-operators` | READ_ONLY, POWER_OPS |
+| `vsphere-admins` | READ_ONLY, POWER_OPS, VM_LIFECYCLE |
+| `vsphere-host-admins` | READ_ONLY, POWER_OPS, VM_LIFECYCLE, HOST_ADMIN |
+| `vsphere-super-admins` | ALL (full access) |
+
+**Security Note:** Users with NO recognized groups are denied ALL access. There is no default permission.
+
+### Tool → Permission Mapping
+
+All 94 tools are mapped to permission levels in `src/mcvsphere/permissions.py`:
 
 ```python
-class CredentialBroker:
+# READ_ONLY - 32 tools
-    """Maps OAuth users to vCenter credentials."""
+"list_vms", "get_vm_info", "list_snapshots", "get_vm_stats", ...
 
-    def __init__(self, vcenter_host: str, fallback_user: str = None, fallback_password: str = None):
+# POWER_OPS - 14 tools
-        self.vcenter_host = vcenter_host
+"power_on", "power_off", "create_snapshot", "revert_to_snapshot", ...
-        self.fallback_user = fallback_user  # Service account fallback
-        self.fallback_password = fallback_password
-        self._connections: dict[str, ServiceInstance] = {}
 
-    def get_connection_for_user(self, oauth_user: dict) -> ServiceInstance:
+# VM_LIFECYCLE - 33 tools
-        """Get pyvmomi connection for this OAuth user."""
+"create_vm", "clone_vm", "delete_vm", "add_disk", "deploy_ovf", ...
-        username = oauth_user.get("preferred_username")
 
-        # Try per-user credentials first
+# HOST_ADMIN - 6 tools
-        try:
+"enter_maintenance_mode", "reboot_host", "shutdown_host", ...
-            vcenter_creds = self._lookup_credentials(username)
-            return self._get_or_create_connection(
-                vcenter_creds["user"],
-                vcenter_creds["password"]
-            )
-        except KeyError:
-            # Fall back to service account
-            if self.fallback_user:
-                return self._get_or_create_connection(
-                    self.fallback_user,
-                    self.fallback_password
-                )
-            raise ValueError(f"No vCenter credentials for user: {username}")
 
-    def _lookup_credentials(self, username: str) -> dict:
+# FULL_ADMIN - 11 tools
-        """Look up vCenter credentials for OAuth user."""
+"run_command_in_guest", "write_guest_file", "restart_service", ...
-        # Option 1: Environment variable
-        env_key = f"VCENTER_PASSWORD_{username.upper().replace('@', '_').replace('.', '_')}"
-        if password := os.environ.get(env_key):
-            return {"user": f"{username}@vsphere.local", "password": password}
-
-        # Option 2: HashiCorp Vault (production)
-        # return vault_client.read(f"secret/vcenter/users/{username}")
-
-        raise KeyError(f"No credentials found for {username}")
 ```
 
 ---
 
-## FastMCP OAuth Integration
+## Implementation Details
 
-### 1. Add OIDCProxy to server.py
+### Key Files
+
+| File | Purpose |
+|------|---------|
+| `src/mcvsphere/auth.py` | OIDCProxy configuration |
+| `src/mcvsphere/permissions.py` | Permission levels and tool mappings |
+| `src/mcvsphere/middleware.py` | RBACMiddleware implementation |
+| `src/mcvsphere/audit.py` | Audit logging with user context |
+| `src/mcvsphere/server.py` | Server setup with OAuth + RBAC |
+
+### RBACMiddleware Flow
 
 ```python
-import os
+class RBACMiddleware(Middleware):
-from fastmcp import FastMCP
+    """Intercepts all tool calls to enforce permissions."""
-from fastmcp.server.auth import OIDCProxy
 
-# Configure OAuth with Authentik
+    async def on_call_tool(self, context, call_next):
-auth = OIDCProxy(
+        # 1. Extract user from OAuth token
-    # Authentik OIDC Discovery URL
+        claims = self._extract_user_from_context(context.fastmcp_context)
-    config_url=os.environ["AUTHENTIK_OIDC_URL"],
+        username = claims.get("preferred_username", "unknown")
-    # e.g., "https://auth.example.com/application/o/vsphere-mcp/.well-known/openid-configuration"
+        groups = claims.get("groups", [])
 
-    # Application credentials from Authentik
+        # 2. Check permission
-    client_id=os.environ["AUTHENTIK_CLIENT_ID"],
+        tool_name = context.message.name
-    client_secret=os.environ["AUTHENTIK_CLIENT_SECRET"],
+        if not check_permission(tool_name, groups):
+            required = get_required_permission(tool_name)
+            audit_permission_denied(tool_name, {...}, required.value)
+            raise PermissionDeniedError(username, tool_name, required)
 
-    # MCP Server base URL (for redirects)
+        # 3. Execute tool with timing
-    base_url=os.environ.get("MCP_BASE_URL", "http://localhost:8000"),
+        start = time.perf_counter()
+        result = await call_next(context)
+        duration_ms = (time.perf_counter() - start) * 1000
 
-    # Token validation
+        # 4. Audit log
-    required_scopes=["openid", "profile", "email"],
+        audit_log(tool_name, {...}, result="success", duration_ms=duration_ms)
-
+        return result
-    # Allow Claude Code localhost redirects
-    allowed_client_redirect_uris=["http://localhost:*", "http://127.0.0.1:*"],
-)
-
-# Create MCP server with OAuth
-mcp = FastMCP(
-    "vSphere MCP Server",
-    auth=auth,
-    # Use Streamable HTTP transport for OAuth
-)
 ```
 
-### 2. Access User Identity in Tools
+### Audit Log Format
 
-```python
+```json
-from fastmcp import Context
+{
+  "timestamp": "2025-12-27T08:15:32.123456+00:00",
+  "user": "ryan@example.com",
+  "groups": ["vsphere-admins", "vsphere-operators"],
+  "tool": "power_on",
+  "args": {"vm_name": "web-server"},
+  "duration_ms": 1234.56,
+  "result": "success"
+}
+```
 
-@mcp.tool()
+Permission denied events:
-async def power_on_vm(ctx: Context, vm_name: str) -> str:
+```json
-    """Power on a virtual machine."""
+{
-    # Get authenticated user from OAuth token
+  "timestamp": "2025-12-27T08:15:32.123456+00:00",
-    user = ctx.request_context.user
+  "user": "guest@example.com",
-    username = user.get("preferred_username", user.get("sub"))
+  "groups": ["vsphere-readers"],
-
+  "tool": "delete_vm",
-    # Get vCenter connection for this user
+  "args": {"vm_name": "web-server"},
-    broker = get_credential_broker()
+  "required_permission": "vm_lifecycle",
-    connection = broker.get_connection_for_user(user)
+  "event": "PERMISSION_DENIED"
-
+}
-    # Execute operation
-    content = connection.RetrieveContent()
-    vm = find_vm(content, vm_name)
-    vm.PowerOnVM_Task()
-
-    # Audit log
-    logger.info(f"User {username} powered on VM {vm_name}")
-
-    return f"VM '{vm_name}' is powering on"
 ```
 
 ---
 
-## MCP Transport: Streamable HTTP
+## Configuration
 
-OAuth requires HTTP transport (not stdio). Use Streamable HTTP:
+### Environment Variables
-
-```python
-# In server.py or via environment
-mcp = FastMCP(
-    "vSphere MCP Server",
-    auth=auth,
-)
-
-def main():
-    # Run with HTTP transport for OAuth support
-    mcp.run(transport="streamable-http", host="0.0.0.0", port=8000)
-```
-
-**Transport characteristics:**
-- Single HTTP endpoint (`/mcp`)
-- POST requests with JSON-RPC body
-- Server-Sent Events (SSE) for streaming responses
-- `Authorization: Bearer <token>` header on every request
-- `Mcp-Session-Id` header for session continuity
-
----
-
-## Environment Variables
 
 ```bash
-# Authentik OIDC
+# ═══════════════════════════════════════════════════════════════
-AUTHENTIK_OIDC_URL=https://auth.example.com/application/o/vsphere-mcp/.well-known/openid-configuration
+# OAuth Configuration
-AUTHENTIK_CLIENT_ID=<from-authentik-application>
+# ═══════════════════════════════════════════════════════════════
-AUTHENTIK_CLIENT_SECRET=<from-authentik-application>
+OAUTH_ENABLED=true
+OAUTH_ISSUER_URL=https://auth.example.com/application/o/mcvsphere/
+OAUTH_CLIENT_ID=<from-oidc-provider>
+OAUTH_CLIENT_SECRET=<from-oidc-provider>
+OAUTH_BASE_URL=https://mcp.example.com  # Public URL for callbacks
+OAUTH_SCOPES='["openid", "profile", "email", "groups"]'
 
-# MCP Server
+# ═══════════════════════════════════════════════════════════════
-MCP_BASE_URL=https://mcp.example.com  # Public URL for OAuth redirects
+# Transport (must be HTTP for OAuth)
+# ═══════════════════════════════════════════════════════════════
 MCP_TRANSPORT=streamable-http
+MCP_HOST=0.0.0.0
+MCP_PORT=8080
 
-# vCenter Connection (service account fallback)
+# ═══════════════════════════════════════════════════════════════
+# vCenter Connection (service account)
+# ═══════════════════════════════════════════════════════════════
 VCENTER_HOST=vcenter.example.com
-VCENTER_USER=svc-mcp@vsphere.local
+VCENTER_USER=svc-mcvsphere@vsphere.local
 VCENTER_PASSWORD=<service-account-password>
-VCENTER_INSECURE=true
+VCENTER_INSECURE=false
 
-# Per-user credentials (optional, for testing)
+# ═══════════════════════════════════════════════════════════════
-VCENTER_PASSWORD_RYAN=<ryan's-vcenter-password>
+# Optional
-VCENTER_PASSWORD_ALICE=<alice's-vcenter-password>
+# ═══════════════════════════════════════════════════════════════
+LOG_LEVEL=INFO
+```
 
-# User Mapping Mode
+### Server Startup Banner
-USER_MAPPING_MODE=service_account  # or 'per_user', 'ldap_sync'
+
+When OAuth is enabled, the server shows:
+```
+mcvsphere v0.2.1
+────────────────────────────────────────
+Starting HTTP transport on 0.0.0.0:8080
+OAuth: ENABLED via https://auth.example.com/application/o/mcvsphere/
+RBAC: ENABLED - permissions enforced via groups
+────────────────────────────────────────
 ```
 
 ---
 
-## Authentik Setup (Quick Reference)
+## OIDC Provider Setup
+
+### Authentik (Recommended)
 
 1. **Create OAuth2/OIDC Provider:**
-   - Name: `vsphere-mcp`
+   - Name: `mcvsphere`
-   - Client Type: Confidential
+   - Client Type: **Confidential**
    - Redirect URIs:
-     - `http://localhost:*/callback`
+     - `http://localhost:*/callback` (for local dev)
      - `https://mcp.example.com/auth/callback`
-   - Scopes: `openid`, `profile`, `email`
+   - Signing Key: Select RS256 certificate
-   - Signing Key: Select or create RS256 key
 
 2. **Create Application:**
-   - Name: `vSphere MCP Server`
+   - Name: `mcvsphere`
-   - Slug: `vsphere-mcp`
+   - Slug: `mcvsphere`
-   - Provider: Select the provider above
+   - Provider: Select provider from step 1
-   - Note the **Client ID** and **Client Secret**
 
-3. **Configure Groups (optional):**
+3. **Create Groups:**
-   - `vsphere-admins` - Full access
+   - `vsphere-readers`
-   - `vsphere-operators` - Limited access
+   - `vsphere-operators`
-   - Groups are included in JWT `groups` claim
+   - `vsphere-admins`
+   - `vsphere-host-admins`
+   - `vsphere-super-admins`
+
+4. **Add Scope Mapping for Groups:**
+   - Ensure `groups` claim is included in tokens
+   - Authentik includes this by default
+
+5. **Note Credentials:**
+   - Copy Client ID and Client Secret
+   - Discovery URL: `https://auth.example.com/application/o/mcvsphere/.well-known/openid-configuration`
+
+### Other Providers
+
+The same pattern works with Keycloak, Auth0, Okta, etc. Key requirements:
+- OIDC Discovery endpoint (`.well-known/openid-configuration`)
+- JWT access tokens (not opaque)
+- `groups` claim in tokens with group names
 
 ---
 
-## OAuth Flow (End-to-End)
+## OAuth Flow
 
 ```
-1. Claude Code connects to MCP Server
+1. Client connects to mcvsphere
-   → GET /mcp
+   → POST /mcp (no auth)
-   → Server returns 401 Unauthorized
+   → Server returns 401 + OAuth metadata URL
-   → WWW-Authenticate header includes OAuth metadata URL
 
-2. Claude Code fetches OAuth metadata
+2. Client initiates OAuth flow
-   → Discovers Authentik authorization URL
+   → Opens browser to OIDC provider
-   → Discovers required scopes
+   → User logs in
+   → Provider redirects with authorization code
 
-3. Claude Code initiates OAuth flow
+3. Client exchanges code for tokens
-   → Opens browser to Authentik login page
+   → POST to provider token endpoint
-   → User enters credentials
+   → Receives JWT access token
-   → Authentik redirects back with authorization code
 
-4. Claude Code exchanges code for tokens
+4. Client reconnects with token
-   → POST to Authentik token endpoint
-   → Receives JWT access token + refresh token
-
-5. Claude Code reconnects with Bearer token
    → POST /mcp with Authorization: Bearer <jwt>
-   → Server validates JWT via Authentik JWKS
+   → Server validates JWT via JWKS
-   → Server extracts user identity
+   → Server extracts user + groups
-   → User can now invoke tools
+   → RBACMiddleware checks permissions
+   → User can invoke allowed tools
 
-6. Tool invocation
+5. Tool invocation
-   → Client: "power on web-server VM"
+   → Client: "power on web-server"
-   → Server: Validates token, maps user to vCenter creds
+   → Middleware: Validate user has POWER_OPS
-   → Server: Executes pyvmomi call
+   → Tool: Execute pyvmomi call
-   → Server: Logs "User ryan@example.com powered on web-server"
+   → Audit: Log with user identity
-   → Client: Receives success response
+   → Client: Receive response
 ```
 
 ---
 
-## Implementation Checklist
+## Implementation Status
 
-### Phase 1: Prepare Server
+### Completed
-- [ ] Add `fastmcp[auth]` to dependencies
-- [ ] Create `auth.py` with OIDCProxy configuration
-- [ ] Create `credential_broker.py` for user mapping
-- [ ] Add audit logging to all tools
-- [ ] Update server.py to use HTTP transport
 
-### Phase 2: Deploy Authentik
+- [x] OIDCProxy configuration (`auth.py`)
-- [ ] Docker Compose for Authentik
+- [x] Permission levels and tool mappings (`permissions.py`)
-- [ ] Create OIDC provider and application
+- [x] RBACMiddleware with FastMCP integration (`middleware.py`)
-- [ ] Configure redirect URIs
+- [x] Audit logging with user context (`audit.py`)
-- [ ] Note client credentials
+- [x] Server integration with OAuth + RBAC (`server.py`)
-- [ ] Test OIDC flow manually with curl
+- [x] Startup banner showing OAuth/RBAC status
+- [x] Security fix: deny-by-default for no groups
+- [x] Authentik setup with 5 vsphere-* groups
 
-### Phase 3: Integration
+### Future Enhancements
-- [ ] Configure environment variables
-- [ ] Test with `fastmcp dev` (OAuth mode)
-- [ ] Test with Claude Code (`claude mcp add --auth oauth`)
-- [ ] Verify audit logs show correct user identity
 
-### Phase 4: Production
+- [ ] Per-user vCenter credential mapping (Vault integration)
-- [ ] HTTPS via Caddy reverse proxy
+- [ ] Rate limiting per user
-- [ ] Secrets in Docker secrets / Vault
+- [ ] Session management and token refresh
-- [ ] Service account with minimal vCenter permissions
+- [ ] Admin tools for permission management
-- [ ] Log aggregation and monitoring
+- [ ] Prometheus metrics for RBAC decisions
 
 ---
 
-## Files to Create/Modify
+## Security Considerations
 
-```
+1. **Default Deny**: Users without recognized groups get NO access
-esxi-mcp-server/
+2. **Token Validation**: JWTs validated via OIDC provider's JWKS endpoint
-├── src/esxi_mcp_server/
+3. **Audit Trail**: All operations logged with user identity
-│   ├── auth.py                 # NEW: OIDCProxy configuration
+4. **Secrets**: Client secrets should be stored securely (env vars, Docker secrets, Vault)
-│   ├── credential_broker.py    # NEW: OAuth → vCenter credential mapping
+5. **HTTPS**: Production deployments should use TLS (via Caddy, nginx, etc.)
-│   ├── server.py               # MODIFY: Add auth, HTTP transport
+6. **Service Account**: Use minimal vCenter permissions for the service account
-│   └── mixins/
-│       └── *.py                # MODIFY: Add ctx.request_context.user logging
-├── .env.example                # MODIFY: Add OAuth variables
-└── docker-compose.yml          # MODIFY: Add Authentik services
-```
 
 ---
 
-## Key Insight
+## Troubleshooting
 
-The "middleman" role of the MCP server is critical:
+### "401 Unauthorized" on all requests
+- Check `OAUTH_ISSUER_URL` points to valid OIDC discovery endpoint
+- Verify client ID and secret match provider configuration
+- Ensure token hasn't expired
 
-```
+### "Permission denied" errors
-OAuth Token (Authentik) ──┐
+- Check user's group memberships in OIDC provider
-                          │
+- Verify groups claim is included in JWT (decode at jwt.io)
-                          ▼
+- Confirm group names match exactly (e.g., `vsphere-admins` not `vsphere_admins`)
-                    ┌─────────────┐
-                    │  MCP Server │  ← Validates OAuth, maps to vCenter creds
-                    │ (Middleman) │  ← Logs audit trail with OAuth identity
-                    └─────────────┘
-                          │
-                          ▼
-              vCenter API (pyvmomi)
-```
 
-The MCP server doesn't pass OAuth tokens to vCenter. Instead, it:
+### Token validation fails
-1. **Authenticates** users via OAuth (trusts Authentik)
+- Ensure OIDC provider issues JWTs (not opaque tokens)
-2. **Authorizes** by mapping OAuth identity to vCenter credentials
+- Check signing key is configured in provider
-3. **Audits** by logging all actions with the OAuth user identity
+- Verify `OAUTH_BASE_URL` matches redirect URI in provider
-4. **Executes** vCenter API calls using mapped credentials
 
-This gives you SSO-like experience while working within vCenter 7.0.3's authentication limitations.
+### Audit logs not showing user
+- Check `groups` scope is requested
+- Verify token contains `preferred_username` or `email` claim

README.md

@@ -195,11 +195,41 @@ Claude: [snapshots 12 VMs in parallel]
 | `VCENTER_CLUSTER` | Target cluster | *auto-detect* |
 | `VCENTER_DATASTORE` | Default datastore | *auto-detect* |
 | `VCENTER_NETWORK` | Default network | *auto-detect* |
-| `MCP_TRANSPORT` | `stdio` or `sse` | `stdio` |
+| `MCP_TRANSPORT` | `stdio` or `streamable-http` | `stdio` |
 | `LOG_LEVEL` | Logging verbosity | `INFO` |
 
 ---
 
+## Multi-User / OAuth Mode
+
+For shared infrastructure or production deployments, mcvsphere supports OAuth 2.1 with any OIDC provider (Authentik, Keycloak, Auth0, etc.):
+
+```bash
+# Enable HTTP transport with OAuth
+export MCP_TRANSPORT=streamable-http
+export OAUTH_ENABLED=true
+export OAUTH_ISSUER_URL=https://auth.example.com/application/o/mcvsphere/
+export OAUTH_CLIENT_ID=your-client-id
+export OAUTH_CLIENT_SECRET=your-client-secret
+export OAUTH_BASE_URL=https://mcp.example.com
+
+uvx mcvsphere
+```
+
+Users authenticate via browser, and group memberships map to permission levels:
+
+| Group | Access |
+|-------|--------|
+| `vsphere-super-admins` | Full control (all 94 tools) |
+| `vsphere-host-admins` | Host operations + VM management |
+| `vsphere-admins` | VM lifecycle management |
+| `vsphere-operators` | Power ops + snapshots |
+| `vsphere-readers` | Read-only |
+
+See [OAUTH-ARCHITECTURE.md](OAUTH-ARCHITECTURE.md) for detailed setup instructions.
+
+---
+
 ## Docker
 
 ```bash

README_DOCKER.md

@@ -235,15 +235,89 @@ MCP_LOG_LEVEL=DEBUG
 log_level: "DEBUG"
 ```
 
+## OAuth Multi-User Mode
+
+For shared infrastructure or production deployments, mcvsphere supports OAuth 2.1 with any OIDC provider. This enables:
+
+- **Browser-based authentication** via Authentik, Keycloak, Auth0, Okta, etc.
+- **Group-based RBAC** with 5 permission levels
+- **Audit logging** with user identity
+
+### Quick Start (Existing OIDC Provider)
+
+If you already have an OIDC provider:
+
+```bash
+# 1. Copy and configure environment
+cp .env.oauth.example .env
+# Edit .env with your OIDC provider details
+
+# 2. Start mcvsphere with OAuth
+docker compose -f docker-compose.oauth-standalone.yml up -d
+
+# 3. Add to Claude Code
+claude mcp add -t http vsphere https://mcp.example.com/mcp
+```
+
+### Quick Start (New Authentik Deployment)
+
+To deploy mcvsphere with a complete Authentik identity provider:
+
+```bash
+# 1. Generate secrets and configure
+./scripts/setup-oauth.sh
+
+# 2. Start full stack (mcvsphere + Authentik + PostgreSQL + Redis)
+docker compose -f docker-compose.oauth.yml up -d
+
+# 3. Configure Authentik at https://auth.yourdomain.com
+#    - Create OAuth2 provider
+#    - Create vsphere-* groups
+#    - Assign users to groups
+```
+
+### RBAC Permission Groups
+
+Create these groups in your OIDC provider:
+
+| Group | Access Level |
+|-------|--------------|
+| `vsphere-super-admins` | Full control (all 94 tools) |
+| `vsphere-host-admins` | Host operations + VM management |
+| `vsphere-admins` | VM lifecycle management |
+| `vsphere-operators` | Power ops + snapshots |
+| `vsphere-readers` | Read-only |
+
+Users without any `vsphere-*` group are denied access (default-deny security).
+
+### Required Environment Variables
+
+```bash
+# Transport (must be streamable-http for OAuth)
+MCP_TRANSPORT=streamable-http
+
+# OIDC Provider
+OAUTH_ENABLED=true
+OAUTH_ISSUER_URL=https://auth.example.com/application/o/mcvsphere/
+OAUTH_CLIENT_ID=your-client-id
+OAUTH_CLIENT_SECRET=your-client-secret
+OAUTH_BASE_URL=https://mcp.example.com
+```
+
+See [OAUTH-ARCHITECTURE.md](OAUTH-ARCHITECTURE.md) for detailed setup and troubleshooting.
+
+---
+
 ## Production Deployment
 
 ### Security Recommendations
 
-1. Use a dedicated user account for vCenter access
+1. Use a dedicated service account for vCenter access
-2. Enable API key authentication
+2. Enable OAuth authentication (not API keys)
-3. Use valid SSL certificates (set `insecure: false`)
+3. Use valid SSL certificates (set `VCENTER_INSECURE=false`)
 4. Limit container resources
 5. Use Docker secrets for sensitive data
+6. Deploy behind a reverse proxy with HTTPS (Caddy recommended)
 
 ### High Availability
 

docker-compose.oauth-standalone.yml

@@ -0,0 +1,64 @@
+# mcvsphere with OAuth - Standalone Mode
+# For users who already have an OIDC provider (Authentik, Keycloak, Auth0, Okta, etc.)
+#
+# Usage:
+#   1. Copy .env.oauth.example to .env
+#   2. Configure your OIDC provider settings
+#   3. docker compose -f docker-compose.oauth-standalone.yml up -d
+#
+# Requires:
+#   - External OIDC provider with OAuth 2.1 support
+#   - Caddy network (caddy-docker-proxy) for HTTPS termination
+#
+# For full Authentik deployment, use docker-compose.oauth.yml instead
+
+services:
+  mcvsphere:
+    build:
+      context: .
+      dockerfile: Dockerfile
+    image: mcvsphere:${MCVSPHERE_VERSION:-latest}
+    container_name: mcvsphere
+    restart: unless-stopped
+    env_file:
+      - .env
+    environment:
+      # Transport - streamable-http required for OAuth
+      MCP_TRANSPORT: streamable-http
+      MCP_HOST: 0.0.0.0
+      MCP_PORT: 8080
+      # OAuth - set in .env file
+      OAUTH_ENABLED: ${OAUTH_ENABLED:-true}
+    volumes:
+      - ./logs:/app/logs
+    networks:
+      - caddy
+    healthcheck:
+      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8080/.well-known/oauth-authorization-server')"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 15s
+    labels:
+      # Caddy reverse proxy - configure domain in .env
+      caddy: ${MCP_DOMAIN:-mcp.localhost}
+      caddy.reverse_proxy: "{{upstreams 8080}}"
+      # TLS - use 'internal' for local dev, remove for production (auto HTTPS)
+      caddy.tls: ${MCP_TLS_MODE:-internal}
+      # WebSocket/streaming support
+      caddy.reverse_proxy.flush_interval: "-1"
+      caddy.reverse_proxy.transport: http
+      caddy.reverse_proxy.transport.read_timeout: "0"
+      caddy.reverse_proxy.transport.write_timeout: "0"
+    deploy:
+      resources:
+        limits:
+          memory: 512M
+          cpus: '1.0'
+        reservations:
+          memory: 256M
+          cpus: '0.25'
+
+networks:
+  caddy:
+    external: true

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "mcvsphere"
-version = "0.2.1"
+version = "0.2.2"
 description = "Model Control for vSphere - AI-driven VMware virtual machine management via MCP"
 readme = "README.md"
 license = "MIT"

src/mcvsphere/middleware.py

@@ -3,12 +3,18 @@
 Provides decorators and hooks for wrapping tool execution with:
 - OAuth permission validation
 - Audit logging with user identity
+- FastMCP middleware integration for RBAC
 """
 
+import logging
 import time
 from collections.abc import Callable
 from functools import wraps
-from typing import Any
+from typing import TYPE_CHECKING, Any
+
+import mcp.types as mt
+from fastmcp.server.middleware import CallNext, Middleware, MiddlewareContext
+from fastmcp.tools.tool import ToolResult
 
 from mcvsphere.audit import (
     audit_log,
@@ -24,6 +30,11 @@ from mcvsphere.permissions import (
     get_required_permission,
 )
 
+if TYPE_CHECKING:
+    from fastmcp.server.context import Context
+
+logger = logging.getLogger(__name__)
+
 
 def with_permission_check(tool_name: str) -> Callable:
     """Decorator factory for permission checking and audit logging.
@@ -81,24 +92,26 @@ def with_permission_check(tool_name: str) -> Callable:
 def extract_user_from_context(ctx) -> dict[str, Any] | None:
     """Extract user information from FastMCP context.
 
+    Uses FastMCP's dependency injection to get the access token from
+    the current request's context variable.
+
     Args:
-        ctx: FastMCP Context object.
+        ctx: FastMCP Context object (unused, kept for API compatibility).
 
     Returns:
         User info dict from OAuth token claims, or None if not authenticated.
     """
-    if ctx is None:
-        return None
-
-    # Try to get access token from context
     try:
-        # FastMCP stores the access token in request_context
+        # FastMCP stores access token in a context variable, accessed via dependency
-        if hasattr(ctx, "request_context") and ctx.request_context:
+        from fastmcp.server.dependencies import get_access_token
-            token = getattr(ctx.request_context, "access_token", None)
+
-            if token and hasattr(token, "claims"):
+        access_token = get_access_token()
-                return token.claims
+        if access_token and hasattr(access_token, "claims"):
-    except Exception:
+            return access_token.claims
-        pass
+    except (RuntimeError, ImportError) as e:
+        # RuntimeError: No active HTTP request context
+        # ImportError: FastMCP auth dependencies not available
+        logger.debug("Could not get access token: %s", e)
 
     return None
 
@@ -179,3 +192,154 @@ def get_permission_summary() -> dict[str, list[str]]:
         summary[level].sort()
 
     return summary
+
+
+class RBACMiddleware(Middleware):
+    """FastMCP middleware for Role-Based Access Control.
+
+    Integrates with FastMCP's middleware system to enforce permissions
+    on every tool call based on OAuth group memberships.
+
+    Example:
+        mcp = FastMCP("my-server", auth=oauth_provider)
+        mcp.add_middleware(RBACMiddleware())
+    """
+
+    def _extract_user_from_context(
+        self, fastmcp_ctx: "Context | None"
+    ) -> dict[str, Any] | None:
+        """Extract user claims from FastMCP context.
+
+        Uses FastMCP's dependency injection to retrieve the access token
+        from the current request's context variable.
+
+        Args:
+            fastmcp_ctx: FastMCP Context object (unused, kept for API compatibility).
+
+        Returns:
+            User claims dict from OAuth token, or None if not authenticated.
+        """
+        try:
+            # FastMCP stores access token in a context variable, accessed via dependency
+            from fastmcp.server.dependencies import get_access_token
+
+            access_token = get_access_token()
+            if access_token and hasattr(access_token, "claims"):
+                return access_token.claims
+        except (RuntimeError, ImportError) as e:
+            # RuntimeError: No active HTTP request context
+            # ImportError: FastMCP auth dependencies not available
+            logger.debug("Could not get access token: %s", e)
+
+        return None
+
+    def _get_username(self, claims: dict[str, Any] | None) -> str:
+        """Extract username from OAuth claims.
+
+        Args:
+            claims: OAuth token claims dict.
+
+        Returns:
+            Username string, or 'anonymous' if no claims.
+        """
+        if not claims:
+            return "anonymous"
+
+        for claim in ("preferred_username", "email", "sub"):
+            if value := claims.get(claim):
+                return str(value)
+
+        return "unknown"
+
+    def _get_groups(self, claims: dict[str, Any] | None) -> list[str]:
+        """Extract groups from OAuth claims.
+
+        Args:
+            claims: OAuth token claims dict.
+
+        Returns:
+            List of group names, or empty list if no claims.
+        """
+        if not claims:
+            return []
+
+        groups = claims.get("groups", [])
+        if isinstance(groups, list):
+            return groups
+        return []
+
+    async def on_call_tool(
+        self,
+        context: MiddlewareContext[mt.CallToolRequestParams],
+        call_next: CallNext[mt.CallToolRequestParams, ToolResult],
+    ) -> ToolResult:
+        """Intercept tool calls to enforce RBAC permissions.
+
+        Args:
+            context: Middleware context containing tool call params.
+            call_next: Next handler in the middleware chain.
+
+        Returns:
+            Tool result if permitted.
+
+        Raises:
+            PermissionDeniedError: If user lacks required permission.
+        """
+        # Extract tool name and arguments from the request
+        tool_name = context.message.name
+        tool_args = context.message.arguments or {}
+
+        # Get user info from OAuth context
+        claims = self._extract_user_from_context(context.fastmcp_context)
+        username = self._get_username(claims)
+        groups = self._get_groups(claims)
+
+        # Set up audit context for this request
+        set_current_user(claims)
+
+        # Check permission
+        if not check_permission(tool_name, groups):
+            required = get_required_permission(tool_name)
+            logger.warning(
+                "Permission denied: user=%s groups=%s tool=%s required=%s",
+                username,
+                groups,
+                tool_name,
+                required.value,
+            )
+            audit_permission_denied(tool_name, tool_args, required.value)
+            raise PermissionDeniedError(username, tool_name, required)
+
+        # Permission granted - execute tool with timing
+        start_time = time.perf_counter()
+        try:
+            result = await call_next(context)
+            duration_ms = (time.perf_counter() - start_time) * 1000
+
+            # Audit successful execution
+            # ToolResult can be complex, just log that it succeeded
+            audit_log(tool_name, tool_args, result="success", duration_ms=duration_ms)
+
+            logger.debug(
+                "Tool executed: user=%s tool=%s duration=%.2fms",
+                username,
+                tool_name,
+                duration_ms,
+            )
+
+            return result
+
+        except PermissionDeniedError:
+            # Re-raise without additional logging
+            raise
+        except Exception as e:
+            duration_ms = (time.perf_counter() - start_time) * 1000
+            audit_log(tool_name, tool_args, error=str(e), duration_ms=duration_ms)
+            logger.error(
+                "Tool failed: user=%s tool=%s error=%s duration=%.2fms",
+                username,
+                tool_name,
+                str(e),
+                duration_ms,
+            )
+            raise

src/mcvsphere/permissions.py

@@ -120,10 +120,11 @@ TOOL_PERMISSIONS: dict[str, PermissionLevel] = {
     "configure_ntp": PermissionLevel.HOST_ADMIN,
     "set_service_policy": PermissionLevel.HOST_ADMIN,
     # ═══════════════════════════════════════════════════════════════════════
-    # FULL_ADMIN - Everything including guest OS and service control (10 tools)
+    # FULL_ADMIN - Everything including guest OS and service control (11 tools)
     # ═══════════════════════════════════════════════════════════════════════
     "start_service": PermissionLevel.FULL_ADMIN,
     "stop_service": PermissionLevel.FULL_ADMIN,
+    "restart_service": PermissionLevel.FULL_ADMIN,
     # Guest OS operations (requires guest credentials, high privilege)
     "run_command_in_guest": PermissionLevel.FULL_ADMIN,
     "list_guest_processes": PermissionLevel.FULL_ADMIN,
@@ -191,9 +192,10 @@ def get_user_permissions(groups: list[str] | None) -> set[PermissionLevel]:
 
     Returns:
         Set of granted permission levels (union of all group permissions).
+        Returns empty set if no recognized groups (deny all access).
     """
     if not groups:
-        return {PermissionLevel.READ_ONLY}
+        return set()  # No groups = no permissions (enforces RBAC)
 
     permissions: set[PermissionLevel] = set()
 
@@ -201,10 +203,7 @@ def get_user_permissions(groups: list[str] | None) -> set[PermissionLevel]:
         if group in GROUP_PERMISSIONS:
             permissions.update(GROUP_PERMISSIONS[group])
 
-    # Default to read-only if no recognized groups
+    # No fallback - unrecognized groups get no permissions
-    if not permissions:
-        permissions.add(PermissionLevel.READ_ONLY)
-
     return permissions
 
 

src/mcvsphere/server.py

@@ -9,6 +9,7 @@ from fastmcp import FastMCP
 from mcvsphere.auth import create_auth_provider
 from mcvsphere.config import Settings, get_settings
 from mcvsphere.connection import VMwareConnection
+from mcvsphere.middleware import RBACMiddleware
 from mcvsphere.mixins import (
     ConsoleMixin,
     DiskManagementMixin,
@@ -68,6 +69,11 @@ def create_server(settings: Settings | None = None) -> FastMCP:
         auth=auth,
     )
 
+    # Add RBAC middleware when OAuth is enabled
+    if settings.oauth_enabled:
+        mcp.add_middleware(RBACMiddleware())
+        logger.info("RBAC middleware enabled - permissions enforced via OAuth groups")
+
     # Create shared VMware connection
     logger.info("Connecting to VMware vCenter/ESXi...")
     conn = VMwareConnection(settings)
@@ -137,8 +143,9 @@ def run_server(config_path: Path | None = None) -> None:
         )
         if settings.oauth_enabled:
             print(f"OAuth: ENABLED via {settings.oauth_issuer_url}", file=sys.stderr)
+            print("RBAC: ENABLED - permissions enforced via groups", file=sys.stderr)
         else:
-            print("OAuth: disabled", file=sys.stderr)
+            print("OAuth: disabled (single-user mode)", file=sys.stderr)
         print("─" * 40, file=sys.stderr)
 
     # Create and run server