Open-source developer tools

Tools for
production systems

Observability, testing, evals and cloud infrastructure. Every tool is built for production and every claim is backed by a benchmark.

flightrec: stdout
// POST /checkout -> 200. Three records were buffered, then discarded.
{"level":"INFO","msg":"request completed","flightrec.dropped_records":3,"status":200}

// POST /checkout -> 502. The error is written at once, followed by everything that led up to it.
{"level":"ERROR","msg":"payment provider timed out","attempt":3}
{"level":"DEBUG","msg":"loaded cart","flightrec.flush_reason":"error_log","items":3}
{"level":"DEBUG","msg":"reserved inventory","flightrec.flush_reason":"error_log"}
{"level":"INFO","msg":"calling payment provider","flightrec.flush_reason":"error_log"}

What we build

Projects

Small, focused tools with real documentation, tests and published benchmarks.

flightrec

Released

A flight recorder for Go's log/slog. Full debug detail for requests that fail or run slowly, one summary line for the ones that don't.

GoobservabilityApache-2.0

GlobalStack

In development

Local Azure development and testing. Runs a bounded Azure management API on your machine, with explicit compatibility boundaries, so infrastructure can be tested without creating cloud resources.

TypeScriptAzureinfrastructure

Evals and benchmarks

In development

Tooling for evaluating infrastructure code and AI agents, and for measuring performance claims reproducibly.

evalsbenchmarkstesting

How we work

Principles

Measured, not claimed

Performance numbers come from benchmark runs anyone can repeat, with the method published next to the result.

Safe by default

Secure defaults first, with CodeQL, OpenSSF Scorecard and fuzzing running in CI.

Honest boundaries

Every project documents what it does not do, so you can decide before you adopt it.

Open source

Permissive licenses, public issue trackers and changelogs that list breaking changes first.