Open Source AI Tools
We build tools for AI development workflows. Four of them — Orunla, CGC, AgentMesh, and Claude LabBook — are free and open source. They were built to solve real problems we encountered while developing our AI SEO platform.
These are developer tools, not part of our agency services. We open-sourced them because they're useful to anyone working with AI coding assistants and we believe the AI developer community benefits from more open tooling.
Orunla
Persistent AI Memory System
Orunla gives AI coding assistants persistent memory across sessions. AI agents like Claude Code, Cursor, and ChatGPT forget everything between conversations. Orunla provides a knowledge graph that remembers facts, relationships, and context — so your AI can pick up where it left off.
Key Features
- Local-first knowledge graph with SQLite storage
- MCP (Model Context Protocol) server for Claude Code and Cursor integration
- CLI tool for terminal-based usage
- REST API for custom integrations
- Temporal decay — memories fade naturally if not accessed
- On-device ONNX-based embeddings (no external API calls for semantic search)
- Built in Rust for speed and low resource usage
How We Use It
We use Orunla in every AI-assisted development session. It stores project architecture, decisions, debugging insights, and user preferences. When we start a new session, the AI already knows the project context. It's the tool that makes our own AI development workflow efficient.
CGC
Context Graph Connector
CGC connects AI agents to your data — files, databases, spreadsheets — without burning through context window tokens. Instead of dumping entire files into the AI's context, CGC maps the data structure first, lets you sample selectively, and chunks large files into manageable pieces.
Key Features
- Multi-source data connectivity (filesystem, databases, spreadsheets)
- Schema discovery and structural mapping
- Smart chunking for large files
- Data sampling without full reads
- MCP server for Claude Code and Cursor integration
- Session management for tracking work across conversations
- Search within data sources without loading full content
How We Use It
CGC is how we keep AI development context-efficient. When exploring a new codebase or data set, we use CGC to discover the structure, sample key files, and only pull in the specific sections we need. It reduces context window waste significantly, leading to better AI responses with less token usage.
AgentMesh
Agent-to-Agent Messaging
AgentMesh lets AI coding agents talk to each other. When you have multiple AI agents working on separate projects — a frontend agent and a backend agent, for instance — AgentMesh enables them to collaborate and share knowledge in real time via MCP, without manual copy-pasting between sessions.
Key Features
- Multi-agent communication across different AI tools and projects
- Automatic message routing through a central broker
- LLM-powered proxy responses when target agents are offline
- Real-time notifications via inbox watcher
- MCP integration for Claude Code, Cursor, and other MCP-compatible tools
- Node.js and Python client libraries included
How We Use It
We use AgentMesh when running multiple Claude Code sessions on different parts of a project simultaneously. A frontend agent can ask the backend agent about available API endpoints, or a testing agent can query the implementation agent about expected behavior — all without us manually relaying information between windows.
Claude LabBook
Experiment Tracking for AI Agents
Claude LabBook gives AI coding agents persistent experiment tracking and semantic code search. AI agents lose context during long sessions — when context windows compress or new conversations start, they repeat failed attempts, revert working code, and rediscover dead ends. LabBook prevents that by logging every trial, decision, and outcome.
Key Features
- Session tracking for organizing related problem-solving attempts
- Trial logging capturing code changes with outcomes and learnings
- Decision logging with architectural rationale
- Pre-change verification surfacing prior trials before modifying components
- Semantic code search indexed by functionality, not just keywords
- MCP server for Claude Code and Cursor integration
- Auto-generated briefing file for instant context recovery
How We Use It
LabBook runs in every development session alongside Orunla and CGC. It's the tool that prevents our AI agents from going in circles. When a context window compresses mid-session, a single briefing call restores the full history of what was tried, what worked, and what failed — so the agent picks up where it left off instead of starting from scratch.
Open Source and Our Agency Work
Orunla, CGC, AgentMesh, and Claude LabBook are developer tools, not part of our AI SEO services for agencies. Our agency work uses a separate set of proprietary tools (BotSight, the query generator, the multi-platform scraper, and the GEO Knowledge Base) that are purpose-built for AI search optimization.
We open-source what benefits the broader developer community. We keep proprietary what gives our agency clients a competitive advantage.
Need AI SEO for Your Agency's Clients?
These open-source tools are for developers. Our agency services use a separate set of proprietary tools built specifically for AI search optimization. Start with a free check to see where your client stands.