Frigate

Overview

Frigate is an open-source NVR that pairs IP-camera recording with AI object detection on local hardware. It is best for camera owners who are comfortable managing a Docker host and want footage kept inside the home.

Key features

Frigate supports 24/7 recording as well as retention rules based on detected objects. That gives users a choice between a complete record and storage tied to activity, rather than motion alone. With a supported AI accelerator, it can distinguish objects such as people and cars from motion, which can cut down on irrelevant alerts. The maker says compatible accelerators can run more than 100 detections per second, but that performance depends on the hardware.

Real-time object tracking and zones let users tune notifications to precise locations in a camera view. This is useful when activity in one area matters more than movement elsewhere. Frigate also supports real-time alerts and behavior and event detection. Integrations include Home Assistant, OpenHab, NodeRed and systems using MQTT. The Home Assistant integration exposes camera, image, sensor, switch and binary sensor entities, and supports browsing snapshots and recordings.

An HTTP API covers cameras, classification, events, media and recordings, making Frigate suitable for systems that need to connect to its data. The maker says all processing is local and camera feeds stay in the home, a meaningful advantage for owners who do not want footage sent elsewhere. Frigate stores user information in its database and hashes passwords with PBKDF2-SHA256 using 600,000 iterations. It does not natively support OIDC, SAML or LDAP, so organizations that require those protocols must use an upstream authentication proxy.

Pricing

The free Frigate plan costs nothing and includes the open-source NVR running as a Docker container on a Docker host. It is the natural starting point for self-hosters who can provide and maintain their own deployment. There is no paid Frigate core plan described; the optional Frigate+ Subscription costs $50 a year, billed annually. It adds access to Frigate+ models, up to 12 fine-tuned models and AI-suggested labels for uploads, while supporting open-source Frigate. That subscription is aimed at users who want those model capabilities; it does not remove the Docker-host requirement. Frigate+ training supports specific labels only, with no custom labels yet.

Platforms

Frigate is available through API, Linux, self-hosted deployment and the web. It runs in Docker on Docker hosts, including as a Home Assistant App. Documentation recommends Docker Compose and says it runs best with Docker on bare-metal Debian-based distributions. Windows is not officially supported, although some users have run it under WSL or VirtualBox; Windows-first users should expect an unsupported setup rather than a standard installation path.

Who it's for

Frigate is a strong fit for technically capable home-camera owners who want local AI detection, configurable retention and Home Assistant or MQTT integration. Its reliance on Docker and the value of supported AI acceleration make it less suitable for people seeking a turnkey Windows install or who do not want to manage their own host. Organizations requiring native OIDC, SAML or LDAP should also account for the proxy requirement.

Pros and cons

• Pros: Local processing keeps camera feeds in the home, according to the maker, which suits privacy-conscious owners.
• Pros: Object-based retention, continuous recording, zones and real-time alerts provide meaningful control over what is recorded and when notifications fire.
• Pros: Home Assistant entities, MQTT support and an HTTP API offer several ways to connect Frigate to a broader setup.
• Cons: Docker is required, and the recommended bare-metal Debian-based setup makes deployment a poor match for users who want a conventional desktop installer.
• Cons: Windows has no official support, and native OIDC, SAML and LDAP are absent, adding work for some users and organizations.
• Cons: Frigate+ model training is restricted to supported labels, so teams needing custom labels cannot use it for that purpose yet.

Alternatives

For broader AI video analytics options, browse AI Video Analytics Software. Roboflow is worth considering when model training and inference credits are the priority: its free tier includes 10 credits monthly, described as enough for about 30 model trainings or 80,000 inferences, while Frigate centers on local camera recording and detection. Amazon Rekognition Content Moderation is an API and web option with a 12-month free tier that includes 60 video-analysis minutes per month; choose it when that hosted free allowance suits the workload better than a self-hosted NVR.

Camio offers usage-based service levels matched to needs, with custom pricing. Scylla Command Center is an enterprise option with custom pricing and support for 100+ video-wall displays, multi-site management and redundancy features. Spot AI has a free dashboard tier with limited analysis volume and no hardware or support, plus paid services priced by order form. Triya AI is another paid option for Android, iOS, self-hosted and web environments. Videonetics AI Video Analytics has a 30-day free trial and custom pricing. Lumeo starts its Core plan at $99 a month, with 10 streams and 30-day recording retention; it may suit readers who want a hosted stream-based service rather than managing Frigate's Docker deployment.

Verdict

Choose Frigate if you want an open-source, locally processed camera NVR and are willing to run Docker; its recording controls, object-based detection and home-automation integrations make that setup worthwhile. Look elsewhere if you need official Windows support, native enterprise identity protocols or a managed deployment that avoids maintaining a host.