OpenCode X
OpenCode X is a terminal AI coding agent under MIT, forked from opencode with 280+ commits. It matches Claude Code's autonomous goals, background execution, parallel subagents and hooks, and adds LLM compression of tool outputs, a three-tier context safety net, doom-loop detection, and support for 75+ model providers. Existing Claude Code hooks, plugins, agents and skills work without changes.
What does OpenCode X do?
OpenCode X is a terminal AI coding agent forked from opencode, per its README, with 280 additional commits at the time the README was read. The README says it matches Claude Code's autonomous goals (/goal with a judge model and a 200-turn cap), background execution (Leader+D to detach a running session), parallel subagents (a Task tool and a swarm mode), a hooks system that shares Claude Code's events and environment variables, worktree isolation, persistent memory, and a skills system compatible with ~/.claude/skills/.
Its README lists three areas where it differs from Claude Code: LLM compression of tool outputs before they reach the main model (three strategies, EXTRACT, SUMMARIZE and FILTER, picked automatically); a three-tier context safety net (a tool-result budget, a MicroCompact pass at 75 percent, and a Context Collapse fallback at 97 percent, plus a reactive 413 retry); and provider freedom through the Vercel AI SDK, with 75+ providers named.
Key facts
- Licence: MIT, per the LICENSE file. Copyright 2025 opencode, since OpenCode X is a fork of opencode.
- Install:
npm install -g @sdeonvacation/opencode-x, or a signed binary from GitHub Releases for macOS (arm64 and x64), Linux (glibc and musl arm64 and x64), and Windows (arm64 and x64). - Providers: any of 75+ providers reachable via the Vercel AI SDK, per the README. Set an API key such as
ANTHROPIC_API_KEY,OPENAI_API_KEY,GOOGLE_API_KEY, and runopencode-x. - LLM compression uses one of three strategies per tool output: EXTRACT for grep, glob and bash; SUMMARIZE for read and prose; FILTER for logs and diffs. Falls back to raw output on error, per the README.
- Three-tier context safety: Tool Result Budget (always on, 50K char cap on the oldest tool output), MicroCompact at 75 percent used, Context Collapse at 97 percent used (writes an emergency backup to disk plus a structured summary). Reactive 413 retry follows.
- Autonomous goals:
/goal <condition>with a separate judge model, a 200-turn hard limit, and an optional token budget. - Swarm execution: one tool call spawns N parallel subagents (1 to 20 concurrent, up to 128 items), with per-item failure isolation.
- Claude Code hooks in
~/.claude/settings.jsonwork automatically, per the README. - Built with Effect-TS and TypeScript 5.8.
What does it replace, and where does it fall short?
OpenCode X replaces Claude Code for someone who wants the same terminal workflow on any model provider, paid to that provider directly. Existing Claude Code hooks and skills keep working, per the README. It is a candidate on the open-source alternatives to Claude Code page.
Where it falls short: the README lists what Claude Code has that OpenCode X does not yet, including agent teams with peer-to-peer communication and shared task lists, dynamic JS orchestration scripts with up to 1000 agents per run, /batch for auto-splitting codebase changes across worktree-isolated PRs, desktop and IDE surfaces (VS Code, JetBrains, web), an agent-view dashboard, and Anthropic's managed background daemon. Other terminal options include Crush and Codex.
How does OpenCode X run?
The recommended install is npm: npm install -g @sdeonvacation/opencode-x, then opencode-x in a terminal. Prebuilt single binaries are on GitHub Releases for macOS, Linux and Windows. A build from source needs Bun and runs ./packages/opencode/script/build.ts --single.
The README says any API key works: ANTHROPIC_API_KEY, OPENAI_API_KEY, GOOGLE_API_KEY, or any of 75+ providers reachable through the Vercel AI SDK. There is no account and no hosted paid version. Existing Claude Code settings and hooks in ~/.claude/settings.json are read automatically.
Who is OpenCode X for?
A developer who already uses Claude Code and wants a drop-in replacement that reads the same hooks and skills, on any model provider, with tighter control over context and cost. A team that needs the multi-surface product (desktop, IDE, web) or Anthropic's managed background daemon should stay on Claude Code.
What limits does the README state?
The README names each Claude Code feature that OpenCode X does not yet match: agent teams with peer-to-peer communication and shared task lists, dynamic workflows with up to 1000 agents per run and 16 concurrent, /batch for splitting changes into worktree-isolated PRs, multi-surface (desktop app, VS Code, JetBrains, web interface), an agent-view dashboard for visual monitoring, and managed background infrastructure (claude agents). LLM compression falls back to raw output on error, per the README, so a failure there does not crash the run.
Questions people ask
Is OpenCode X open source?
Yes. The LICENSE file is MIT, an OSI approved licence with no additional restriction. The copyright is held by opencode, the upstream project OpenCode X forks. Anyone can read, modify and redistribute the code under MIT terms, including for commercial use.
Does OpenCode X work with Claude Code hooks and skills?
Yes. The README says existing Claude Code hooks in ~/.claude/settings.json, plugins, agents, skills and CLAUDE.md files all work in OpenCode X without changes, and it shares the hooks system's events and environment variable names. That is the migration path the README puts first.
Which models can OpenCode X use?
Any of 75+ providers reachable through the Vercel AI SDK, per the README. Named examples include Claude, GPT, Gemini, Mistral, Llama and DeepSeek. You bring your own API key and pay the provider directly. There is no subscription. The README also describes a hybrid routing mode that sends compression and title jobs to a cheap model and keeps reasoning on the premium one.
What is the three-tier context safety net?
A three-layer set of safeguards that stop a session from crashing on a token overflow, per the README. The first tier is a per-tool-output cap (50K characters on the oldest output, always on). The second is a MicroCompact pass at 75 percent context used, summarising older messages and keeping the ten most recent verbatim. The third is a Context Collapse at 97 percent, writing an emergency backup to disk with a structured summary. A reactive 413 retry runs the same turn inline after the emergency compaction.
How is OpenCode X different from opencode?
OpenCode X is a fork of opencode with 280 additional commits, per its README. The additions are named: LLM compression, cache stability, doom loop detection, orchestration guardrails (spawn depth 3, descendant cap 50, per-model concurrency semaphore), swarm execution, autonomous goals with judge validation and a 200-turn cap, background execution with Leader+D, hybrid model routing, and native compatibility with Claude Code hooks and skills. Everything else is inherited from opencode.
Sources
- OpenCode X README and LICENSE: github.com/sdeonvacation/opencode-x, read 2026-09-28. The README describes the fork's differences from opencode and Claude Code in detail.
More coding assistants and agents
An MIT terminal coding agent with a full-access build agent and a read-only plan agent switched with the Tab key, installers for every platform, a beta desktop app, and the most-starred repository in this directory.
OpenAI's Apache-2.0 coding agent that runs locally in your terminal, with a desktop app and IDE extensions, signed in with a ChatGPT plan or an OpenAI API key, so the code is open and the models are OpenAI's.
Google's Apache-2.0 terminal agent for Gemini models with a free tier of 1,000 requests a day on a personal Google account, built-in search grounding, file and shell tools, MCP support, checkpointing and a GitHub Action.
A high-performance code editor from the creators of Atom and Tree-sitter, GPL-3.0-or-later with Apache-2.0 components, with an agent panel, external agents over ACP, edit prediction, MCP tools, sandboxing, and local models through Ollama or llama.cpp.
An MIT self-hosted control centre, called Agent Canvas, that runs the OpenHands agent or Claude Code, Codex, Gemini and any ACP-compatible agent across local, Docker and cloud backends, with automations triggered by schedules and webhooks.
An Apache-2.0 coding agent for VS Code, JetBrains, the terminal and a desktop app that edits files as reviewable diffs, runs commands with your approval, works with any model provider including Ollama, and connects tools through MCP.
Added September 28, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 28, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/sdeonvacation/opencode-x.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.