
UniFi Protect MCP
io.github.sirkirbyPublished by GitHub account sirkirbyv0.1.6Updated Oct 26, 2026
Manage UniFi Protect cameras, events, recordings, and smart detections via MCP.
Overview
Local MCP server that lets an assistant query and manage UniFi Protect cameras, events, recordings, smart detections and devices on your own controller.
- What you can do
- List, inspect, snapshot and control cameras, including PTZ, settings, recording toggle and reboot
- Query historical events, smart detections, Find Anything searches and thumbnails, with real-time event streaming
- Check recordings and export clips with timelapse support
- Manage Known Faces groups and inspect liveviews
- Control lights, sensors and chimes, and read NVR health, firmware and viewer status
- Manage Alarm Manager rules, with AI-powered alarms visible only to SuperAdmin accounts
- When to use it
- Useful when you run UniFi Protect on a local controller and want to ask an assistant in natural language about camera activity, detections, recordings or device health, or to make occasional camera and light changes with a preview-then-confirm step.
- Requirements
- Runs locally over stdio, installed from PyPI (uvx or pip) or Docker. Needs network access to the controller plus UNIFI_HOST (or UNIFI_PROTECT_HOST), UNIFI_USERNAME and UNIFI_PASSWORD; optional UNIFI_PORT, UNIFI_VERIFY_SSL and UNIFI_API_KEY for some HDR, microphone and light settings. No .env file is loaded automatically.
Installation
In SourceWeft
- Open UniFi Protect MCP in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
UniFi Protect MCP Server
MCP server exposing UniFi Protect tools for AI assistants and other MCP-capable clients. Query cameras, events, smart detections, Find Anything detection search, recordings, lights, sensors, chimes, Known Faces, license plates, and the Alarm Manager with safe-by-default permissions and preview-before-confirm for all mutations.
Software integrating over HTTP, including automation services and MCP adapters,
should use unifi-api-server for REST and GraphQL reads, actions,
and SSE events.
Install
Claude Code (recommended)
The plugin installs the MCP server, an agent skill for tool discovery, and a guided setup command:
Then run the interactive setup to configure your controller connection:
This walks you through entering your controller host, credentials, and permission preferences. Credentials are given as a file or command provider, never typed into chat. Setup saves the result as the plugin's Claude Code options, which apply in every project; a password or API key entered as an option is stored in the system keychain. Start a new Claude Code session after setup to connect.
Upgrading from a plugin version before 1.0.0, which kept settings in the project's .claude/settings.local.json? Follow the plugin upgrade guide.
Codex
Register the marketplace, then install unifi-protect from Codex's /plugins UI:
After installing, ask Codex to use the UniFi Protect setup skill. Setup saves a version-pinned MCP server entry with your controller settings in Codex's config.toml and prompts you to restart Codex. After each plugin upgrade, re-run setup with --refresh.
PyPI / Docker
Usage Examples
Once connected, just ask your AI agent in natural language:
"List all cameras that detected motion in the last hour"
"Show me smart detection events from the front door camera today — people and vehicles only"
"Find driveway detections for white vans this week"
"Which cameras have the most motion events this week? Any unusual patterns?"
"Are there any cameras offline or with degraded connections?"
"Show me all recording events from the driveway camera between 2 AM and 5 AM last night"
"What sensors triggered alerts today and what were the readings?"
All camera and event queries are read-only by default. Mutations (camera settings, light controls) use a preview-then-confirm flow.
Configure
Set these variables in the server's process environment. The plugin setup skill already does this for Codex and OpenClaw; in Claude Code the plugin maps its options onto these variables. The server does not automatically load .env or working-directory YAML files; load a trusted env file explicitly in your launcher (Docker env_file: is supported), or select custom YAML with an absolute CONFIG_PATH. See configuration for migration examples.
With no controller host configured, the server still starts in setup mode: it tells the client how to configure it, offers only the read-only protect_setup_status tool, and makes no controller connections.
Camera HDR and microphone volume, plus floodlight power, brightness, PIR sensitivity,
and duration now use the Protect public Integration API with uiprotect 16.x.
Set UNIFI_PROTECT_API_KEY (or the shared fallback UNIFI_API_KEY) and restart
the server before using these settings. Session credentials remain required for
other Protect operations. Missing keys or incompatible public device IDs fail
before any settings in the request are written.
Microphone volume updates accept 1–100; zero is rejected by the public API.
Existing HDR aliases are preserved: true/on/normal mean auto, false
means off, and always/superHdr select public HDR on. Light brightness
remains 1–6, sensitivity 0–100, and duration 15–900 seconds. light_on and
is_light_on are equivalent inputs and control forced illumination; turning
forced illumination off does not disable motion-triggered lighting.
Settings updates retain preview/confirmation and policy gates. Requests are validated before writing; controller failures can still leave a partial update. Failed updates return an error with the settings that applied, so callers can inspect the current state before retrying.
Fallback: The shared UNIFI_* variables (e.g., UNIFI_HOST) also work. The server checks for UNIFI_PROTECT_* first and falls back to UNIFI_* if the server-specific variable is not set. For single-controller setups, the shared variables are all you need.
MCP response size
For tool results that already provide structured output, adaptive is the default response mode. It classifies each request by the canonical date-based protocolVersion advertised during MCP initialization, not by the client's product name or application version. Requests advertising MCP 2025-06-18 or later receive concise content plus the full result once in structuredContent; requests advertising an earlier revision (such as 2024-11-05 or 2025-03-26), or whose revision metadata is missing or malformed, keep full compatibility JSON in content. Set UNIFI_PROTECT_MCP_CONTENT_MODE to compat to force the duplicated compatibility form, or to compact to force concise text plus full structured output outside a negotiated request. This server-specific variable overrides UNIFI_MCP_CONTENT_MODE; use compat for any client that consumes the full result only from content, regardless of its advertised revision.
The lazy-loading meta-tools (*_tool_index, *_execute, *_batch, *_batch_status, and lazy-only *_load_tools) remain content-only; they are not the pre-2025-06-18 protocol category described above. For structured inner results, *_execute and *_batch_status expose one normalized JSON payload in content rather than a nested transport pair; content-only execute results remain unchanged. Response modes do not convert these meta-tools to structuredContent.
When Network is enabled alongside Protect, its largest source responses are independently bounded: unifi_get_dashboard defaults to summary=true, and unifi_list_rogue_aps defaults to a summarized page of at most 100 records; summary=false restores the full selected data.
AI-powered alarms need SuperAdmin. The alarm-rule tools (
protect_alarm_list_rules/protect_alarm_get_rule/protect_alarm_create_rule/protect_alarm_update_rule/protect_alarm_delete_rule) transparently use the modern UniFi-OS Alarm Manager when the account is SuperAdmin, and fall back to the classic automations view otherwise. The modern path surfaces and manages AI-powered alarms (e.g. AI Natural Language); the legacy path cannot see those rules and responses include a standard MCP_metanotice when the view is limited. Grant the account SuperAdmin on the console hosting Protect to view/manage AI alarms. Blast radius: on a standalone UNVR this is contained to Protect; on a combined UDM console SuperAdmin also grants Network/UniFi-OS control.
Sensitive response fields
Protect tools redact secret-bearing stream fields by default before returning data to MCP clients. This includes RTSP/RTSPS stream aliases and URLs from protect_get_camera_streams. Disable redaction for a trusted local administration process with UNIFI_PROTECT_REDACT_SENSITIVE_FIELDS=false or the global UNIFI_REDACT_SENSITIVE_FIELDS=false policy flag when raw stream values are required.
Run
Claude Desktop
Add to claude_desktop_config.json:
Features
- Cameras -- list, inspect, snapshot, RTSP streams, PTZ control, settings, recording toggle, reboot
- Events -- query historical events, smart detections (person/vehicle/animal/package), Find Anything detection search, thumbnails
- Real-time streaming -- websocket event buffer with MCP resource subscriptions and polling
- Recordings -- status, availability, clip export with timelapse support
- Known Faces -- list, rename, merge, and remove face recognition groups
- Devices -- lights (brightness, PIR sensitivity), sensors (temperature, humidity, motion), chimes (volume, trigger)
- Liveviews -- list and inspect multi-camera layouts
- System -- NVR info, health metrics, firmware status, connected viewers
Agent Skills
The Protect plugin ships one agent skill that works alongside the MCP tools:
Security Digest
Cross-product event intelligence that generates a concise security summary across all connected UniFi systems.
- Sources: Protect camera events, Access door events, Network firewall activity
- Severity classification: Events are classified by time-of-day context (business hours vs. after hours vs. overnight) and event type (person detection, vehicle, door access, firewall block, etc.)
- Cross-product correlation: Five built-in correlation rules surface patterns that span products:
- Motion at a door camera without a corresponding badge-in
- Multiple failed door access attempts in a short window
- Person detection coinciding with firewall blocks from the same timeframe
- After-hours camera activity with no Access event
- Repeated vehicle detections at the perimeter
- Activity counts: Aggregated totals across all sources for quick at-a-glance awareness
Invoke via the skill command after installing the plugin:
Event Improvements
camera_name in Event Responses
All event-related tools now include camera_name alongside camera_id in every event object. The name is resolved from bootstrap data cached at startup — no extra API calls required.
Affected tools: protect_list_events, protect_list_smart_detections, protect_recent_events
Before:
After:
This eliminates the need to call protect_list_cameras separately just to map IDs to names.
Compact Mode
protect_list_events and protect_list_smart_detections accept a compact=true parameter that strips low-signal fields from responses: thumbnail_id, category, sub_category, and is_favorite. This produces responses roughly 40% smaller.
Compact mode is the recommended default when building summaries or feeding events into downstream prompts where token budget matters.
Documentation
- Configuration -- Full env var reference, YAML config, Protect-specific options
- Permissions -- Permission system, category defaults, how to enable mutations
- Tool Catalog -- All 63 tools organized by category
- Event Streaming -- Real-time event architecture, MCP resources, polling
- Troubleshooting and support bundles -- Reviewed support evidence, connection issues, SSL, missing tools
Development
See the root CONTRIBUTING.md for the full monorepo workflow.
License
Source: apps/protect/README.md at commit fbd083c
Tools
0Version history
1- v0.1.6LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts