Claude Code is brilliant. But it doesn't know your organization.
Ask it why your team chose JWT over sessions—blank. Ask who has expertise in worker thread optimization—it guesses. Ask what failed the last time someone tried this approach—it hallucinates.
Today we're releasing gutt-claude-code-plugin—an open-source plugin that gives Claude Code organizational memory.
Two commands:
/plugin marketplace add iBrain-BVBA/gutt-claude-code-plugin
/plugin install gutt-claude-code-plugin@gutt-plugins
Your AI assistant now knows your organization.
What "Knowing" Actually Means
Not search. Not retrieval. Understanding through relationships.
When we say Claude Code "knows" your organization, we mean it has access to structured knowledge about:
- Who has expertise in what domains
- What decisions were made and the reasoning behind them
- Where patterns, conventions, and implementations live
- How similar problems were solved—and what failed
This isn't keyword matching against a document dump. It's querying gutt—where facts connect to other facts, decisions link to their rationale, and context compounds over time.
The difference is the difference between a shoebox and a photo album.
Shoebox: You dump documents in. You search keywords. You get fragments back. No relationships. No "why." No temporal awareness of what's still relevant.
Album: Organized. Captioned. Connected. You see how things relate. You understand context, not just content. You know when something was true and whether it still is.
gutt builds the album. Claude Code reads it.
How It Works
The plugin connects Claude Code to gutt—organizational memory for AI agents.
Before starting work, Claude Code queries gutt:
fetch_lessons_learned(query="authentication patterns")
search_memory_nodes(query="rate limiting", entity_types=["Decision", "Lesson"])
search_memory_facts(query="Maria HAS_EXPERTISE_IN")
It retrieves relevant context—lessons learned, architectural decisions, team expertise—before generating a single line of code.
After completing work, it captures new knowledge:
add_memory(
name="Lesson: Worker thread memory leak resolution",
episode_body="Root cause was unbounded event listener accumulation...",
source="development_session"
)
Decisions get linked to reasons. Solutions get linked to problems. gutt grows smarter with every session.
What Gets Stored
gutt captures relationships as triplets:
"JWT auth" CHOSEN_OVER "sessions" BECAUSE "stateless API requirement"
"Maria" HAS_EXPERTISE_IN "worker thread optimization"
"Rate limiting" IMPLEMENTED_IN "middleware/rateLimit.ts"
"Memory leak fix" RESOLVED_BY "Maria" ON "2026-01-15"
These aren't just searchable text. They're queryable relationships with temporal metadata. gutt knows when something became true and surfaces information based on relevance, not just keyword matches.
What's Included
The plugin ships with 10 production-ready hooks:
| Hook | What It Does |
|---|---|
| Session start | Auto-fetches relevant organizational context |
| User prompt submit | Reminds to check memory before starting work |
| Subagent planning | Reviews lessons before execution |
| Task completion | Captures lessons for future sessions |
| Stop | Prompts lesson capture after significant work |
Plus: MCP setup wizard, config discovery, cross-repo memory sync, and post-tool formatting hooks.
Everything runs automatically. Install once, benefit continuously.
The Numbers
We've been dogfooding this plugin while building gutt.
3-5x reduction in repetitive explanations. That's sessions that start productive instead of sessions that start with "let me explain the architecture again."
The compounding effect is harder to measure but more significant. Lessons captured in Week 1 prevented mistakes in Week 4. Decisions documented in one project informed approaches in the next. Context stopped living exclusively in developers' heads.
170% velocity increase during our most intensive development period—despite operating at 66% team capacity.
The plugin doesn't write better code. It prevents the context loss that forces you to write the same explanations over and over.
Why This Matters
AI coding assistants are everywhere now. The productivity gains are real. But there's a ceiling.
That ceiling is context.
Every time you start a new Claude Code session, you're teaching it from scratch. Every developer on your team runs isolated AI assistants that don't share what they've learned. Every insight from yesterday's debugging session evaporates overnight.
Your AI assistants have working memory. They don't have organizational memory.
This creates a strange inversion: the tool that's supposed to make you faster also makes you repeat yourself more. The assistant that generates code in seconds requires minutes of context-setting first.
gutt-claude-code-plugin bridges that gap. Your AI assistants learn from your organization's accumulated knowledge. They know what was tried before. They understand why certain decisions were made. They can query expertise across your entire team.
The result isn't just faster coding. It's AI that actually understands the environment it's coding in.
Open Source
The plugin is fully open source under MIT license.
GitHub: github.com/iBrain-BVBA/gutt-claude-code-plugin
Docs: gutt.pro/docs/claude-code-plugin
We're actively developing and welcome contributions, feedback, and feature requests. This is version 1.0 of our vision for how organizational memory should integrate with development workflows.
Get Started
/plugin marketplace add iBrain-BVBA/gutt-claude-code-plugin
/plugin install gutt-claude-code-plugin@gutt-plugins
Run the setup wizard to connect to your gutt instance. Configuration takes less than a minute.
Don't have gutt yet? Get in touch—we'll get you set up.
Start building AI that knows your organization—not just your syntax.
gutt builds context infrastructure for AI agents. We believe the gap between AI capability and AI usefulness is organizational memory. This plugin is our first step toward closing it.