Occontext Thinking Trim

Name: Occontext Thinking Trim
Author: old-mikser

v1.0.3Observability

@old-mikser/occontext-thinking-trim

OpenCode plugin to trim thinking/reasoning from context to reduce token usage

Repository archivedNo commits in six months

#opencode #opencode-plugin #thinking #reasoning #context #tokens

GitHub repo npm page Homepage

GitHub stars

Monthly installs

11 in 7 days

Composite scoreSCORE

6.7

Multi-signal model

Last commit

6 months ago

2026-02-18

Install and configure

opencode.json

Writes to this project's opencode.json — applies to this repository only.

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["@old-mikser/occontext-thinking-trim@1.0.3"]
}

Writes to ~/.config/opencode/opencode.json — applies to every project.

~/.config/opencode/opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["@old-mikser/occontext-thinking-trim@1.0.3"]
}

opencode loads npm dependencies through its embedded runtime on startup and caches them locally — no manual global install needed.

README.md

View original→

OpenCode plugin that trims thinking/reasoning from context while preserving display.

Why?

Models with extended thinking (Claude, DeepSeek, etc.) send reasoning back in context on each request. This can:

Increase token usage
Potentially influence model behavior in unwanted ways

This plugin removes reasoning from the context sent to the model, while keeping it visible in the UI.

Installation

Add to your opencode.json:

{
  "plugins": ["@old-mikser/occontext-thinking-trim"]
}

On first run, the plugin creates ~/.config/opencode/OCTT.jsonc with default settings.

Configuration

Edit ~/.config/opencode/OCTT.jsonc:

{
  // Enable or disable thinking trim
  "enabled": true,
  
  // Number of recent turns to preserve thinking
  // 0 = trim all thinking from context
  // 2 = keep thinking in last 2 assistant messages
  "keepTurns": 0
}

Options

Option	Type	Default	Description
`enabled`	boolean	`true`	Enable/disable the plugin
`keepTurns`	number	`0`	Number of recent turns to preserve thinking. `0` = trim all

Example

With keepTurns: 2, the last 2 assistant messages will keep their reasoning intact, while older reasoning is removed.

License

MIT