authentik
The first patch on 2026.8 is twenty-odd fixes and no new surface.
A side-by-side editorial comparison of werf and Windmill — release velocity, themes, recent moves, and the top alternatives to consider.
The 3.x line finally ships function: authenticated secret writes and render patches.
werf publishes the same work across alpha, beta, ea and stable channels plus a 3.x dev line, so one fix surfaces three or four times under different version numbers. This pair is that pattern exactly: v3.3.0 carries two features — writing authenticated secret values from deploy, and renderPatches support — plus four fixes, and v2.77.2 carries the same four fixes and nothing else. The fixes are the usual build-engine territory: buildah imports failing on symlinked paths, a panic when a stapel base image disappears mid-commit, a cache repo pointing at itself, and host-cleanup misreporting what it pruned.
Windmill gives its AI agents a test harness, the piece that lets them into production flows
Windmill is a developer platform for scripts, flows and internal apps, shipping weekly across three lanes. The AI lane is the busiest: sessions that build scripts and flows as reviewable drafts, a read-only plan mode that refuses writes until you approve, version history on session artifacts, and now evaluation for reusable agents. The runtime lane keeps absorbing data tooling — dbt projects run natively, BigQuery and Snowflake came out from behind the Enterprise licence. The workspace lane is consolidating deployment semantics around fork lineage.
werf publishes the same work across alpha, beta, ea and stable channels plus a 3.x dev line, so one fix surfaces three or four times under different version numbers. This pair is that pattern exactly: v3.3.0 carries two features — writing authenticated secret values from deploy, and renderPatches support — plus four fixes, and v2.77.2 carries the same four fixes and nothing else. The fixes are the usual build-engine territory: buildah imports failing on symlinked paths, a panic when a stapel base image disappears mid-commit, a cache repo pointing at itself, and host-cleanup misreporting what it pruned.
The division of labour between the lines is now explicit — 2.x takes only fixes, 3.x takes fixes plus everything new. This is the first window where the 3.x dev line carries user-facing deploy function rather than build internals and gated experiments, which moves it from a place to accumulate changes toward something with its own reason to exist. The build-engine hardening under parallelism continues underneath both.
renderPatches and authenticated secret writes are both deploy-side, so the next 3.x releases likely continue there rather than returning to buildah. Whether the embedded deno work behind its build tag connects to the render path is still not stated.
Windmill is a developer platform for scripts, flows and internal apps, shipping weekly across three lanes. The AI lane is the busiest: sessions that build scripts and flows as reviewable drafts, a read-only plan mode that refuses writes until you approve, version history on session artifacts, and now evaluation for reusable agents. The runtime lane keeps absorbing data tooling — dbt projects run natively, BigQuery and Snowflake came out from behind the Enterprise licence. The workspace lane is consolidating deployment semantics around fork lineage.
The AI work has moved through a recognisable sequence: first let the assistant build things, then constrain what it may do without approval, then make its output reviewable and revertible, and now make agents measurable against a fixed set of cases. That is the arc of taking AI from a demo surface to something you are willing to put in a scheduled flow. In parallel, Windmill keeps pulling data-stack runtimes inside its own execution model rather than integrating with them, and keeps giving away runtimes that were previously licensed.
Evals are the obvious hook for regression gating, so expect them to be callable from a flow or CI rather than only from the agent card, and expect the eval run history to feed the same version-comparison surface the session artifacts already use.
Other Infra & APIs products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either werf or Windmill.
The first patch on 2026.8 is twenty-odd fixes and no new surface.
Alert rate limits stop being per-source and start being per-tenant, by query.
The essay series keeps running ahead of the products it describes.
Both lines patch on the same day, and the reverted JSONata upgrade finally sticks.
Folder-level RBAC lands, and approvals finally get a webhook to talk to.
Retool posts one-line changelog entries, and its agent work is the only visible thread
See all werf alternatives → · See all Windmill alternatives →
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
They serve adjacent needs but don't currently overlap on shipped themes. Windmill is currently shipping more aggressively (velocity 6.3 vs 5.0), with 1 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. Windmill is currently shipping more aggressively (velocity 6.3 vs 5.0), with 1 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other Infra & APIs products to evaluate alongside.
Top werf alternatives in Infra & APIs are ranked by recent ship velocity. Browse the "werf alternatives" section above for the current picks, or visit /alternatives/werf for the full list with editorial commentary on each.
Top Windmill alternatives in Infra & APIs are ranked by recent ship velocity. Browse the "Windmill alternatives" section above for the current picks, or visit /alternatives/windmill for the full list with editorial commentary on each.