Claude Sonnet 4.6 is the backbone of our entire platform, Humans.Team. Over 85 development sessions, Claude Code (powered by Sonnet) built 90% of our Next.js application — from Supabase database architecture and Row Level Security policies to AI journal integration, real-time notifications, PWA offline support, and a bilingual FR/EN system across 30+ pages.
What sets Sonnet 4.6 apart is its ability to hold deep context across long sessions. It remembers architectural decisions from hours ago, understands our codebase patterns, and writes production-ready TypeScript that rarely needs fixing. The reasoning is exceptional — it debugs complex issues by tracing through multiple files and connections.
We also use Claude Desktop daily for content strategy, press releases, blog articles, and bilingual copywriting. The nuance in both French and English is remarkable.
Claude Code Routines is a new automation layer inside Claude Code that lets developers configure scheduled or event-triggered AI coding tasks with no cron jobs, no servers, and no infrastructure to maintain.
The problem: Developers already using Claude Code to automate their dev cycle were still managing cron jobs, infrastructure, and MCP server tooling themselves. The AI was capable, but the plumbing was on you.
The solution: A routine is a Claude Code automation you configure once, including a prompt, repo, and connectors, and then run on a schedule, from an API call, or in response to an event. Routines run on Claude Code's web infrastructure, so nothing depends on your laptop being open.
Features worth noting:
Three trigger types in one surface.
Scheduled routines — Triage, label, and post summaries automatically (hourly, nightly, or weekly)
API routines — Each routine gets its own endpoint + auth token; connect to deploy hooks, alerts, and dashboards
GitHub routines — Trigger on PRs via webhooks for reviews, triage, and auto-sync changes
Claude opens one session per PR and continues to feed updates from that PR to the session, so it can address follow-ups like comments and CI failures.
Who it's for: Engineering teams and solo devs who are already on Claude Code and have repetitive workflows they're currently handling with cron jobs, GitHub Actions, or manual scripts. Especially useful for on-call engineers who need faster alert triage and teams maintaining multi-language SDKs.
What repetitive dev workflow would you automate first if you had 15 routine runs a day?
P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends
@rohanrecommends The GitHub routine that keeps one session per PR is the detail that makes this genuinely useful - most tools treat each push as isolated. Does the session retain context across force pushes too, or does that reset things?
@rohanrecommends what does runs on web signify?
Seems like a good idea especially trying to compete with OpenClaw that runs continuously. Could be useful to run cron jobs to scan code databases to increase code quality especially with the volume of code that Claude outputs. If it says what it's going to do then I think its a good addition. Would be nice if it developed a report on what it's learned during those cron sessions and how it will implement those thoughts into deeper levels of model customisation
So Claude Code can now babysit your PRs overnight without your laptop staying open. For on-call engineers drowning in repetitive triage, that's not a small thing. Curious how the 15 daily routine limit holds up for bigger teams in practice. 🤔
I started using Claude Code for repetitive tasks and I'm saving 25 minutes per delivery. That's not a small number when you multiply it across a week. Now Routines is dropping and I'm actually excited — which doesn't happen often.
Love this! I use Claude Code daily to build my security scanner. Routines could be a game changer for repetitive scanning workflows.
I'm saving 25 minutes per delivery on repetitive tasks thanks to Claude Code. I can't wait to see what Routines can do.