Anthropic Blocks Multi-Model OpenCode Plugin, Confirming Lock-In Policy
Anthropic has explicitly blocked the oh-my-openagent plugin (formerly oh-my-opencode), a TypeScript orchestration layer for OpenCode that routes prompts to competing models — Kimi K2.5, GLM-5, GPT-5.4, and others — while preserving Claude Code's UX. The plugin's author claims the block was triggered specifically because the plugin enabled vendor-neutral multi-model use of Claude Code. A hash-anchored edit mechanism in the same tool reportedly raised Grok Code Fast 1 success rate from 6.7% to 68.3%, suggesting the edit-tool layer, not just the model, is where significant leverage sits.
Why It Matters
This is the clearest signal yet that Claude Code's market position is partly defended through access policy rather than pure capability. It has already spawned competing tools (free-claude-code proxy, Kiro IDE) and is accelerating the agent-tools standards consolidation around Agent Skills (agentskills.io) as developers seek harness portability.