TODO for AI
vs OpenCode
Open-source terminal coding agent, bring your own keys, any provider. Here is where each one wins, with numbers, and who should pick which.
Any provider (BYOK)
Every providerWho should pick which
- You want a free MIT tool and have API keys
- You prefer to self-manage every bill and machine
- You want one flat plan covering every model
- You want the machine, browser and vault managed for you
- You need team roles and pooled usage
What OpenCode is, and where it stops
OpenCode is the open-source terminal agent for people who want to own the whole stack. MIT licensed, any provider through your own API keys, a fast TUI, and nothing running anywhere you did not put it. If you already hold keys and like reading source, it is the most hackable option on this list.
You also carry every operational burden. Token bills arrive per provider with no cap, there is no cloud machine, no scheduling, no team roles, and the scope is code. It is a great tool, and a great tool is not a platform.
Developers who want a free, hackable terminal agent and already hold API keys.
Three reasons OpenCode users look elsewhere
Per-token, per-provider, no pooled plan. One runaway loop is a surprise invoice.
Runs on whatever box you are at. Nothing persists, nothing is scheduled.
No roles, approvals or shared usage. Fine for one person, hard for a team.
Six dimensions that decide it
What each one ships
What it costs
Free, MIT. You pay each provider for tokens.
Hobby free (Sonnet only), Starter $20, Pro $100, Ultra $200. Paid plans include every model from every vendor.