Swamp launches today
I’ve been following Swamp closely from the side, so here’s an observer’s note on launch day.
They started in late January, raised a seed round in summer, and today they’re launching. That’s how fast you can go with agent-driven development and a software factory.
It was fast for me too. It clicked almost the next day after the demo at CfgMgmtCamp. I was in the hotel trying stuff with Claude, and it was just working. I immediately used up my $20 subscription and had to upgrade to do more. Later it became my home lab automation. The first thing I solved was DNS for all my devices, with their registration and management in Pi-hole.
You feel the difference most when you don’t have it. At $JOB I didn’t. Instead I got 50+ bash and python scripts that were hard to convince the agent to use, and it just created a new one each time. Without Swamp, agents are clunky and forget everything. With it, what you build stays and gets reused. They call it behavior-centric design.
Speed has a price, though. Let’s put it plainly: they’re a VC-backed startup, and they need to show growth to get investment. Swamp is open source, and the source code is available. For distribution they follow the Red Hat model. At first you didn’t need an account. Later an account was required for advanced features like datastores and vaults. Then at the beginning of October an account became mandatory. After one day of silence, the community reminded them that they had promised a free option for personal use, and the personal license came back. The community held them to their word.
And that’s the part I value most: the people. It’s a great community that talks in its chats, and also writes articles to discuss the complex things they build and discover using Swamp. It’s not only about Swamp. It’s exploration, and a log of practitioners trying things and creating emerging practices for a new era of agent-driven development.