
I have been using OpenCode at work for the past few weeks.
Before OpenCode I was using claude code as my primary harness.
What made me switch to OpenCode was a few reasons
A company wide monthly budget limit per engineer was introduced, which made me more mindful on how I spend the tokens.
I felt Opus and Sonnet was costing me a lot of tokens even to get some simple tasks done.
So I started experimenting with various ZAI open weight models(Kimi K3, GLM 5.3) and OpenAI closed weight models(Terra, Sol, Luna). OpenCode supports switching between models easily.
I also started enjoying the UI of OpenCode better than claude code.
OpenCode shows all the MCPs in the right side bar with the status. So you don’t have to guess which MCP is enabled and available in the session.
Once the MCPs are configured in OpenCode, you can switch between various models easily without having to reconfigure them.
I was starting to get annoyed at the random verbs (Orchestrating, Osmosing, Perambulating, Percolating, Perusing, Philosophising,) that claude code UI uses for thinking.
I have been playing with the Kimi K3 and GLM 5.3 in the past few weeks. Though Impressive I still feel like the OpenAI terra model is better than them.
Also I always start the OpenCode sessions with Herdr. It’s a modern take on Tmux and I enjoy the UI much better.
I recommend playing with OpenCode and Herdr.
