Light Trail

AI Edge

Why shouldn't every agent or app have its own global edge?

To help answer this question we've abstracted two underlying capabilities (our HTTPProxy and WAF) into a single feature that protects and safely routes global internet traffic to backend services.

Key features

  • 100% free and simple to deploy
  • Multiple and custom hostnames
  • Out of the box health checks
  • Full metrics export to Grafana Cloud
  • Optional authentication
  • Powered by the open source Envoy proxy and Coraza WAF
  • Backed by 17+ global points of presence

Next steps

Two Men

Connectors

Safely expose local environments to the internet for free.

We're building for diverse connectivity options, but starting with QUIC-based tunnels and a common use case: safely sharing localhost with the world.

Key features

  • Built with a zero trust architecture
  • Features flexible and lightweight QUIC tunnels
  • Built on Iroh, a library that establishes direct peer-to-peer QUIC connections.
  • Resilient, secure certificate-based routing

Next steps

datumctl

Authenticate and manage resources using a K8s-style syntax (get, apply, delete).

Datum MCP

An official, self-hosted MCP server that is tightly integrated with Cursor.

Developer Tools

Nix, macOS, Windows, and Linux apps for exposing localhost to the internet.

DNS

A secure and resilient authoritative DNS service, served across 17 locations.

Domains

Organize and programmatically manage domain resources across providers.

Metrics Export

Export full OTel metrics to Grafana Cloud along with prebuilt dashboards.

Audit & Activity logs

Production grade infrastructure with built in features for activity / audit logging.

AGPLv3 License

Open source transparency with community contributions and a neutral strategy.

Common Questions

Datum is an early stage company, and our product is evolving steadily to include more core features (e.g. additional Connectors, Galactic VPC, Compute, etc) as well as platform features that support enterprise readiness and operational scale.

Offering a “forever free” service tier is super important to us. As we mature our product, we plan to enable all possible features in the free tier and provide reasonable usage limits. Through this coming year we will introduce a paid plan with an SLA, as well as other deployment models.

Great question! We currently offer a public cloud model (multi-tenant at the control plane and network level) as well as open source. This year, pending demand from design partners, we plan to introduce a managed cloud offering (single tenant control plane) and a BYOC model.

We’ve positioned our product as an open, AI native network cloud. Practically speaking, we offer features that mirror aspects of the large public clouds and CDN’s: an edge proxy, a reverse tunnel, DNS, global reach, a high performance backbone, etc. Aside from being built fresh in 2025, what sets us apart is our target customer (we’re solely focused on modern ‘alt cloud’ providers) and our open, neutral strategy.

The world’s biggest companies and clouds take advantage of traditional internet networking capabilities like peering, interconnection, and deterministic routing. By building out a global physical network and backbone, they are able to understand and control the flow of their traffic, improve performance, lower costs, comply with regulation, and participate with an ecosystem of networks. Our mission is to make these superpowers available to every agent, developer and builder.

Datum’s control plane is built on Kubernetes Custom Resource Definitions (CRDs). You can read some of the background in Scot’s blog here. While you don’t have to use k8s to use Datum Cloud, it will feel quite natural if you do. An important benefit for users is the insane familiarity that LLMs have with the Kubernetes codebase. It sure does help Claude have a more seamless and accurate experience!

Datum deploys infrastructure in the top internet aggregation points globally. These points of presence (PoPs) are packaged into Regions and Availability Zones (AZs). Each Region represents a specific geographic and network boundary, while each AZ provides independent capacity within that Region. Similar to most public clouds, Datum uses a country-anchored naming format with the following elements:

FieldDescription
GeographyTwo-letter ISO country code (example: US, DE, IN)
Cardinal DirectionDirectional indicator within the country: north, south, east, west, or central.
NumberRegion index within that location.
CountAvailability Zone identifier: a, b, c, etc.