OpenCode X CLI
A terminal AI coding agent built on opencode: any model with your own keys, subagent swarms, goals and output compression.
- 78 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The `/goal` and `/swarm` modes write code and run commands on their own for up to 200 turns
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/opencode-xDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The `/goal` and `/swarm` modes write code and run commands on their own for up to 200 turns
This is third-party code. Review the repository files before installing.
What it does
OpenCode X is a fork of opencode with 280 additional commits. The agent works in the terminal with any of 75+ models through the Vercel AI SDK, the user brings their own keys and pays only for tokens. The `/swarm` command launches parallel subagents in batches with bounded concurrency and `/goal` carries work to a goal with judge validation and continuation up to 200 turns. Hooks, plugins, agents, skills and `CLAUDE.md` from Claude Code are picked up without migration. Tool output is compressed by a language model, context has three tiers of overflow protection and there are memory across sessions, push-to-background and doom loop detection. Unused MCP tools are removed through an allowlist.
Who it is for. Developers used to Claude Code who want to pay for tokens with their own keys and choose the model without changing hooks and skills.
Good fit when
- An agent with models from different providers, including local ones, is needed.
- A large batch change is split between parallel subagents with a limit.
- A long task runs to its goal unattended while a judge verifies the result.
- The bill for tool output grows and the output is compressed before it reaches the model.
Not a fit when
- Claude Code features absent here are needed: agent teams, dynamic workflows and a desktop app.
- Official vendor support is required: the project is run by a single author.
- Autonomous goals on code that is not yours without human review are unacceptable.
Example request
/goal Fix the failing tests in the auth module and stop when the judge confirms the result.Limitations
The project is independent and not affiliated with Anthropic or the opencode creators. The comparison with Claude Code in the README is written by the author and does not replace your own check. Your own model provider keys are needed and token prices depend on them. The `/goal` and `/swarm` modes write and change code on their own and the spawn depth and descendant limits are configurable. Some Claude Code features are missing: agent teams, dynamic workflows and desktop apps. The license is MIT with no telemetry and no account.
How to disable. Uninstall the package with `npm uninstall -g @sdeonvacation/opencode-x` and delete the agent settings and session directories. Revoke provider keys with the providers.
FAQ
How do I install it?
With `npm install -g @sdeonvacation/opencode-x`, then run `opencode-x`. Prebuilt binaries for macOS, Linux and Windows exist too.
Is a subscription needed?
No. The agent runs on your provider keys and charges only for tokens used.
What does `/goal` do?
It carries work to a goal, validates it with a judge model and continues automatically up to 200 turns with a token budget.
Will it pick up my Claude Code hooks?
Yes, hooks, plugins, agents and skills work without migration: events and environment variables match.
What about telemetry?
The authors state there is no telemetry and no account.
Discussion
No questions or reviews yet.
Related
CLI toolsAn OpenCode plugin that runs a team of parallel agents with separate sessions, a shared task board and messaging
An open source coding agent for the terminal and desktop with build and plan modes
Orchestrator plugin for OpenCode and Codex CLI: multiple agents, hooks, built-in MCPs, ultrawork mode and hash-anchored edits
Agents, skills and templates for structured AI development in OpenCode, Codex and Claude Code: docs, plans, review, checkpoints
A terminal coding agent built on Codex, tuned for low-cost open models like Kimi K3, DeepSeek and Qwen
The opencode-mcp server hands coding tasks to the OpenCode agent from Claude, Cursor and VS Code and tracks their progress.
README in short
The README compares the agent with Claude Code and opencode, lists capabilities and limits and gives installation by npm, binaries and from source plus a quick start with a provider key.