Coding agent? Read /start for machine-readable setup instructions.
Your whole team clears the queue from the coding agent they already use. It drafts, remembers, and improves — and a human still owns every reply.
Now anyone on your team can do support.
Get started with the agent-operated onboarding. No install step – your agent runs the onboarding coordinator and takes it from there.
fetch https://windmill.sh/start to onboard meCopy + paste this to your coding agent to get started.
No training, no dashboard to learn. The agent already knows your product — it can read your actual code when Windmill lives in your monorepo — remembers every past ticket, and drafts the reply. Your teammate just decides. Holidays covered, founder-tax gone, the next hire trivial.
Customers star-rate the replies they get. Windmill feeds the best and worst conversations back to the agent as examples you curate, and its CLAUDE.md / AGENTS.md playbook sharpens itself. Answer a product question once — the agent knows it forever.
Your institutional knowledge is markdown in your own repo: version-controlled, diffable in a PR, yours forever — not locked inside a SaaS you rent. A human owns every reply and the agent learns what lands, so customers get fast, real answers instead of canned macros.
Tell your agent to run the onboarding command. It scaffolds your support repo — or drops into your existing monorepo so the agent can see the product too.
Your customers file tickets in your hosted portal.
Tell your agent to work the queue. It drafts from your knowledge and history; you decide; it learns.
fetch https://windmill.sh/start to onboard meCopy + paste this to your coding agent to get started.
What's next: customer context and refunds for Stripe and Polar, so billing questions get handled in the same loop.