// the find
DataDog/datadog-agent
Main repository for Datadog Agent
The Datadog monitoring agent — the binary you install on servers to ship metrics, logs, traces, profiles, and security events to Datadog's SaaS platform. It's open source in the sense that you can read and fork it, but it's functionally useless without a Datadog subscription. If you're evaluating it as 'open source observability tooling,' you're looking at the wrong repo.
The eBPF-based system probe does kernel-level network monitoring and runtime security without requiring any changes to application code — it intercepts syscalls directly, which is genuinely hard to do well. OTel compatibility is real and useful: the agent can ingest OpenTelemetry spans and metrics as a collector, so teams standardizing on OTel can still route through it without forking their instrumentation. Multi-language APM auto-instrumentation (Go, Java, Python, .NET, Node, Ruby) works from a single agent install with no code changes in many runtimes. The CI/CD infrastructure is sophisticated — dual GitLab/GitHub pipelines, Bazel for hermetic builds, automated dependency tidying, supply-chain hardening via Chainguard — this is a team that takes release engineering seriously.
It is a data forwarder for a closed paid platform, full stop. There is no self-hosted backend you can point it at. The build system is a three-headed beast: Bazel, Python invoke tasks via dda (Datadog's own CLI), and Go modules all coexist, and the developer docs tell you to install dda before doing anything — contributing a patch is not a casual afternoon. The component-based dependency injection architecture means tracing any behavior requires navigating a significant layer of abstraction before you reach actual logic; the 3.6k star count tells you most people install prebuilt packages and never touch the source. Mixed licensing (Apache 2.0 for userspace, GPL-2.0 for the eBPF code) will trigger enterprise legal review and slow down any internal fork or derivative.