OpenCode
Free, open-source terminal coding agent with 75+ model providers
Overview
OpenCode is an MIT-licensed, terminal-first AI coding agent with build/plan modes, file-snapshot undo, and 75+ supported model providers. Maintained by Anomaly (formerly SST), after a 2025 fork dispute split it from Charm's version, later renamed Crush.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with OpenCode
- MIT-licensed, fully open source
- 75+ supported model providers, including local models
- Plan mode (review changes before they're applied)
- File-snapshot undo for multi-file edits
- TUI, IDE and desktop clients
- OpenCode Go bundled hosted-model subscription
- GitHub Action integration
ASEAN Perspective
OpenCode in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
OpenCode has become the default open-source answer to Claude Code — MIT-licensed, provider-agnostic, and reportedly past 160,000 GitHub stars with millions of monthly users. Plan mode (reason before you let it write) and file snapshots (undo a whole multi-file edit in one shot) are the two features reviewers keep coming back to; they make it usable for real refactors, not just toy demos.
The honest gap versus Claude Code is autonomy on complex, multi-file agentic tasks — reviewers who've used both still give Claude Code the edge there. Setup is also rougher: no one-command install-and-go, more configuration than Cursor or Claude Code expect. Worth knowing: the project split messily from Charm in 2025, and Charm's original fork is now a separate product called Crush — they are not the same tool despite sharing history and a name for a while.
What people say
OpenCode's numbers are the headline: past 160,000 GitHub stars as of early 2026 — reportedly the most-starred open-source coding agent by a wide margin — with figures floating around of several million monthly active developers. Reviewers on sites like Developers Digest and AI Coder Scope frame it explicitly as "the open-source terminal coding agent that challenges Claude Code," which tells you where it's positioned: not a toy, a genuine daily driver for people who don't want to be locked into one vendor's CLI or pay per-seat for Cursor.
The feature reviewers cite most is Plan mode — you get the agent to lay out its intended changes before anything touches disk, which matters on refactors where a wrong first guess wastes a lot of tokens and time. File snapshots get similar praise: a multi-file edit gone wrong is one undo away from a clean state, rather than a git-stash scramble. Provider flexibility (75+ providers, including fully local models) is the other repeated selling point — you're not stuck if your Anthropic quota runs out mid-task.
On the critical side, the consistent line is that Claude Code still wins on autonomous handling of large, multi-file agentic tasks — OpenCode narrows the gap but doesn't close it. Onboarding is also rougher: you configure providers and keys yourself, and the docs assume comfort with terminal tooling; there's no single-command, zero-config path the way Claude Code or Cursor offer.
The project's history is its own review topic. The original repository moved to Charm, which some contributors say involved rewritten git history and deleted comments; a group forked it and kept building under the OpenCode name, which is the project now maintained by the Anomaly org (formerly SST) and covered here. Charm's version was later renamed Crush to end the naming collision. Anyone comparing notes from mid-2025 discussions should check which project a given post is actually describing — GitHub issues and Hacker News threads from that period conflate the two constantly.
OpenCode Go, the optional hosted-model add-on, gets generally positive but thin reviews: $5 for the first month, then $10/month for bundled access to open-weight models like DeepSeek and Qwen, with usage caps ($12/5hrs, $30/week, $60/month) that reviewers say they haven't hit in weeks of daily use, though it's still in beta.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including OpenCode's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with OpenCode unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to OpenCode directly →
Spotted something out of date? Suggest an update →
OpenCode in the news
More in Code & Dev Tools