TDengine is an open-source, cloud-native time-series database written in C and licensed under AGPL-3.0, designed for IoT, connected-car and Industrial IoT workloads that ingest TB and even PB scale data per day from billions of sensors and data collectors.

What it is

TDengine is a high-performance, scalable time-series database and historian aimed at Industrial IoT scenarios. The project is written in C, released under the AGPL-3.0 licence, and its core modules — including the cluster feature and the AI agent — are available under open-source licences. The registry entry shows 25,121 stars, 5,018 forks and 434 open issues, with the most recent push on 17 September 2026. Repository topics include bigdata, cloud-native, cluster, connected-vehicles, database, distributed, financial-analysis, industrial-iot, iot, metrics, monitoring and scalability. Documentation is published in English, Simplified Chinese and Japanese.

The concrete problem it addresses is the operation of time-series pipelines for very large numbers of collection points. TDengine's stated differentiators are solving the high cardinality issue to support billions of data collection points, and outperforming other time-series databases for data ingestion, querying and compression. Rather than assembling a general-purpose time-series store alongside separate caching, stream-processing and subscription components, TDengine bundles caching, stream processing, data subscription and AI agent features into one system, which the README presents as a reduction in system design complexity and operating cost. It lives in the time-series database ecosystem and is positioned by its README against other time-series databases in general terms.

Key capabilities

• Built-in AI agent TDgpt connects to time-series foundation models, large language models, machine-learning and traditional algorithms for forecasting, anomaly detection, imputation and classification; it has its own README at tools/tdgpt/README.md.
• Native distributed design using sharding and partitioning, separation of compute and storage, and RAFT consensus, with Kubernetes deployment support and full observability.
• Super tables, partitioning by time interval and pre-computation support efficient exploration and formatting of time-series data.
• Built-in caching, stream processing and data subscription are included in the database instead of being separate components.
• Targets TB and even PB scale per day of ingestion generated by billions of sensors and data collectors.
• Deployable on public, private or hybrid clouds.
• Repository topics cover financial analysis, metrics and monitoring alongside industrial IoT and connected vehicles.

Who uses it and how

• Industrial IoT and IIoT deployments where large fleets of sensors and data collectors produce continuous telemetry.
• Connected-vehicle programmes, which appear both as a headline use case and as a repository topic.
• Monitoring and metrics platforms, indicated by the metrics and monitoring topics.
• Financial-analysis workloads, indicated by the financial-analysis topic.
• Teams running on public, private or hybrid clouds, including Kubernetes-based cluster deployments.

Getting started

TDengine is installed via container, installation package or Kubernetes, or used as the fully managed TDengine Cloud service without installation; documentation is at docs.tdengine.com. Contributors who want to build, release and test it themselves follow the prerequisites and build sections of the README.

How it compares

The provided facts name no specific competing time-series database, so no product-by-product comparison is possible here; the README contrasts TDengine with "other time-series databases" only in general terms. On that basis it stands alone in this registry as the documented entry for high-cardinality IIoT time-series storage. The comparison axes the source material supports are ingestion, query and compression performance, and the bundling of caching, stream processing, subscription and AI features into a single system.

When to use it — and when not to

A self-hoster takes on a distributed system: sharding and partitioning, separation of compute and storage, RAFT consensus and, on the Kubernetes route, cluster orchestration and observability, so an operations team is required rather than a single process. Teams that cannot accept AGPL-3.0 obligations, or that only need a small single-node metrics store, should look elsewhere. The README is oriented towards contributors building and releasing TDengine from source, with end-user material delegated to the external documentation site, and it cites 23.7k stars while the registry reports 25,121, so the README lags the current state slightly.

sudo apt-get update
sudo apt-get install -y build-essential cmake git

sudo yum groupinstall -y "Development Tools"
sudo yum install -y cmake3 git

git clone https://github.com/taosdata/TDengine.git
cd TDengine
mkdir debug && cd debug
cmake .. -DBUILD_CONTRIB=ON    # First build: -DBUILD_CONTRIB=ON is REQUIRED
make -j$(nproc)