OpenCode is what happens when open-source meets agentic coding — and it's growing fast.
The product: an autonomous coding agent that reads your codebase, plans changes, edits files, runs tests, and reports back. Works with any OpenAI-compatible model — your local Ollama, OpenRouter, Anthropic, DeepSeek.
Why people switch: privacy (your code never leaves your machine with local models), flexibility (switch models freely), and cost (pay per token to your provider of choice, or $0 with local).
The workflow: point it at a repo, describe the task, watch it work. It reads the tree, greps for patterns, edits files, runs the test suite, and iterates until tests pass or it gets stuck.
The rough edges: it can be too eager — always set it up in a branch and review the diff. It also occasionally hallucinates a file path; triple-check before you commit.
Who it's for: developers who want agentic coding without the $200/month subscription and without their code on someone else's server.
Verdict: the most interesting open-source coding agent right now. Try it on a side project first.