ix-opencode-plugin

An OpenCode plugin that brings Ix Memory's graph-first reasoning into OpenCode as a native cognitive layer.

OpenCode + Ix = reasoning engine + persistent code knowledge graph. Skills are cognitive abstractions (not CLI wrappers) that start with cheap graph signals before reading source, and stop early when the question is answered.

Requirements

• OpenCode installed
• Ix Memory CLI (ix) installed and connected to a workspace
• Bun installed (OpenCode's plugin runtime)

opencode --version
command -v ix && ix status
bun --version

Installation

Per-project (recommended)

# From your project root
mkdir -p .opencode/plugins .opencode/tools .opencode/commands .opencode/agents

cp -r /path/to/ix-opencode-plugin/plugins/. .opencode/plugins/
cp -r /path/to/ix-opencode-plugin/tools/.   .opencode/tools/
cp -r /path/to/ix-opencode-plugin/commands/. .opencode/commands/
cp -r /path/to/ix-opencode-plugin/agents/.  .opencode/agents/
cp    /path/to/ix-opencode-plugin/AGENTS.md  .opencode/AGENTS.md

Add to your project's opencode.json:

{
  "plugin": [".opencode/plugins/ix-plugin.ts"],
  "instructions": [".opencode/AGENTS.md"]
}

Global install

cp -r /path/to/ix-opencode-plugin/plugins/. ~/.config/opencode/plugins/
cp -r /path/to/ix-opencode-plugin/tools/.   ~/.config/opencode/tools/
cp -r /path/to/ix-opencode-plugin/commands/. ~/.config/opencode/commands/
cp -r /path/to/ix-opencode-plugin/agents/.  ~/.config/opencode/agents/

Append AGENTS.md content to ~/.config/opencode/AGENTS.md.

Build the graph first

cd /your/project
ix map       # index the codebase (30s–2min)
ix status    # verify graph is present

Skills

Slash commands with phased reasoning — graph first, source reads only when needed.

Command	What it does	Key rule
/ix-understand [target]	Build a mental model of a system or the whole repo	Graph only — no source reads
/ix-investigate <symbol>	Deep dive: what it is, how it connects, execution path	Graph first, one symbol read max
/ix-impact <target>	Change risk: blast radius, affected systems, test targets	Depth scales with risk level
/ix-plan <targets...>	Risk-ordered implementation plan for a set of changes	Parallel impact, finds shared dependents
/ix-debug <symptom>	Root cause analysis from symptom to candidates	Targeted reads at suspects only
/ix-architecture [scope]	Design health: coupling, smells, hotspots	Graph only — never reads source
/ix-docs <target> [--full]	Generate narrative-first system documentation	Default is onboarding-focused

Agents

Agent	Purpose
ix-explorer	General-purpose graph exploration, open-ended questions
ix-system-explorer	Full architectural model of a codebase or region
ix-bug-investigator	Autonomous investigation from symptom to root cause candidates
ix-safe-refactor-planner	Blast radius + safe change sequencing for refactors
ix-architecture-auditor	Full structural health report with ranked improvements

Ask OpenCode to use them directly:

Use ix-explorer to answer: how does the authentication flow work?
Use ix-bug-investigator to find why the session expires early.
Use ix-safe-refactor-planner to plan changes to: UserRepository, AuthService

Tools

Seven CLI-backed tools OpenCode can call during any task:

Tool	Purpose
ix-query	Look up a graph entity by name
ix-neighbors	Traverse the neighborhood of an entity
ix-impact	Blast radius analysis for a target
ix-map	Architectural map of a scope
ix-ingest	Check graph status or trigger a re-index
ix-history	Revision and history lookup
ix-docs-tool	Retrieve doc/context summaries

Automatic hooks

Trigger	Hook	Effect
Any task starts	AGENTS.md context injection	OpenCode knows to use graph data before reading files
Before a file edit	tool.execute.before → ix-pre-edit	Runs impact check, surfaces blast radius
Before a read	tool.execute.before → ix-read	Hints toward graph narrowing first
Before grep/bash	tool.execute.before → ix-intercept	Front-runs with graph search
First run in workspace	plugin init → ix-ingest	Ensures graph is present and fresh
After tool execution	tool.execute.after → ix-errors	Handles stale graph or missing entities

All hooks bail silently if ix is not in PATH or the backend is unreachable.

Further reading

• QUICKSTART.md — step-by-step first use guide
• ARCHITECTURE.md — design decisions and component overview
• TOOL_CONTRACT.md — tool API reference and output format
• AGENTS.md — always-on context injected into every OpenCode session