- Author
- Alexander Belanger, Co-Founder and CEO, Hatchet
- Published
- 3 June 2025
- Kind
- Company blog post with a short code example
- Original
- docs.hatchet.run/blog/go-agents
- Subject
- Agent tools and setups
What it is
Alexander Belanger co-founded Hatchet. The post defines an agent as a looping process. It explains why Go's runtime fits many concurrent agents.
What it says
- Belanger says Hatchet sees hybrid stacks. A Next.js or FastAPI backend pairs with an agent written in Go.
- The post says agents are long-running and expensive per run. They also wait on input and i/o.
- The post says goroutines use little memory. Thousands of concurrent agents can then run on light infrastructure.
- The post says a Go agent loop keeps no state between messages. This allows a durable boundary between messages.
- The post says context.Context gives one cancellation mechanism. It contrasts this with Python and Node.js, where cancelling is hard.
- The post lists drawbacks: weaker third-party support, poor machine learning use, faster rivals.
What evidence it gives
The post offers argument and a simplified Go example of an agent loop. It also relies on the author's observations of Hatchet users. It gives no benchmarks comparing languages. The author admits using mutexes more often than channels in most Go programs.
Interests and limits
Hatchet sells agent orchestration, so the company benefits when teams build agent backends. The post compares languages by reasoning only. Nothing was tested. Claims about Python and Node.js are the author's general view. The post is from June 2025.