YELLA
Automations that don't die.
A durable runtime for event-driven workflows. Crash recovery, at-least-once delivery, effect replay, built-in traces. Open source. Self-host it or we'll run it for a fiver.
Coming soon
Here's how you build on it.
The runtime handles queues, events, retries, crash recovery, and observability. You just describe what you want. Paste the docs into Claude and tell it what to build. It reads the llms-full.txt files, understands the framework, and writes the whole thing.
Copy the starter prompt and give it a go.
Fetch https://yella.tech/llms-full.txt and follow the links. Connect to my Gmail. Every hour, check for new emails. If anything looks urgent, post it to Slack with a summary of why it needs attention. Use existing connectors. If one doesn't exist, create it from the docs. Set up for Yella hosting.
It crashes. It comes back. See everything.
Process dies mid-run? Don't care. It picks up where it left off. API calls that already went through don't go again. Runs, traces, dead letters, connector health: one config line. No Grafana. No Datadog invoice. You sleep through the night.
How durability works →Plugs into everything.
Slack, Telegram, Discord, email, webhooks, cron: all out of the box. Need something else? Describe it in your prompt. Claude will figure it out.
Auth, rate limiting, and reconnection are handled by the framework. Your connector just defines send and receive.
Available connectors →It's SQLite. Relax.
No database server. No connection strings. Ten runs a minute is a rounding error. Need more? Turn a dial.
See the benchmarks →Backed up out of the box.
Every hosted database is WAL-streamed through Litestream. You get a live SQLite database and an up-to-date backup with zero configuration. No managed database, no replication lag, no bill.
Run it yourself, or let us host it.
The framework is free and open source. Don't want to manage infrastructure? Fixed monthly price. No surprises. No metered billing.
30-day free trial on Starter. No card required.