OpenCode
The agent that runs both the free local model and the paid cloud models from one place.
Tested 2026-11-08
OpenCode is the tool I actually sit in front of. It is an open-source coding agent with a terminal interface and a desktop app, and its main advantage is that it does not care which model you point it at. Local Qwen through Ollama and paid DeepSeek side by side, in the same session, switched with one command.
It also does the boring infrastructure well: a permissions system so it asks before it writes or runs something dangerous, project instructions that every session reads first, and skills and MCP servers that give it extra tools without stuffing everything into the prompt.
Pros
- Model-agnostic: the same agent drives free local models and paid ones
- Permissions ask before writes, shell commands and network calls
- Project instruction files, skills and MCP servers without prompt bloat
- Open source and free, with no account required
Cons
- Terminal-first, so the first hour is unfamiliar if you live in an IDE
- You are responsible for model choice, and a bad choice looks like a bad agent
- Configuration lives in files, which is either the best or worst part depending on the day
What I tested
| Test | Result |
|---|---|
| Long document summary | N/A for the app itself. Handed the test document to both a local and a cloud model and compared, which is the point. |
| Coding task | Built a six-page static site from scratch: build script, stylesheet, content file, four tool reviews and two guides. Files written where asked, nothing outside the folder touched. |
| Hard reasoning | N/A. The app routes the question; the model answers it. |
| Image understanding | TODO: test on a screenshot with a vision-capable model. |
| Cost for a week of real use | $0 for OpenCode itself. What it costs is whatever the model behind it costs, which is the right way round. |
Bottom line: The best free way to use one paid model and one local model without maintaining two setups. It is the first thing I install on a new machine.
What this site is
I test AI tools myself — cloud models, local models, and the apps that manage them. Reviews only appear once I’ve run the same prompts through each tool on my own machine or account. No sponsorships, no affiliate links, no ads. More about how this works.
One email a week
What I tested this week, what it cost to run, and the one thing worth your time. No spam, unsubscribe in one click.