Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.selftune.dev/llms.txt

Use this file to discover all available pages before exploring further.

​
Status: Experimental

OpenCode support is in early development. It can capture sessions through a plugin hook adapter, ingest existing sessions, run selftune’s judge and optimizer workflows through the native OpenCode CLI, and use experimental runtime replay validation during selftune evolve.

​
Real-time hooks

selftune opencode install
This installs:
  • A selftune plugin in OpenCode’s auto-discovered plugins directory
  • Bundled selftune optimizer agents in opencode.json

​
Ingest

selftune ingest opencode
Reads OpenCode session logs and imports them into selftune’s database.

​
LLM-backed workflows

OpenCode can run selftune’s LLM-backed workflows directly:
  • Judge and validation calls use opencode run
  • Specialized optimizer agents use opencode run --agent

​
Limitations

  • OpenCode does not expose a prompt-submission hook event, so prompt logging and auto-activate are unavailable
  • OpenCode does not support hard-blocking pre-tool guard exits, so selftune can warn about conflicting writes but cannot enforce them the way Claude Code can
  • Runtime replay validation is still experimental and depends on opencode run --format json
  • This integration is still experimental