A

ai-agents

Production-grade memory operating system for agentic AI. Interoperates with every major framework via MCP, OpenAI-compatible gateway, and native /v1/* REST — Claude Code, OpenClaw, ZeroClaw, Hermes, MemPalace, LangChain, LlamaIndex, CrewAI, AutoGen, Mem0, Letta, Zep. In production use since December 2025. Apache 2.0.

Always-on GitLab repo custodian: an autonomous AI teammate (4 personas) that reviews merge requests, sweeps merged code for regressions, files issues, and drafts fix MRs — tuning itself to your team's standards from your reactions. Built on Vertex AI Gemini + GitLab MCP for the Google Cloud Rapid Agent Hackathon 2026 (GitLab Track).

Desktop kanban board for the markdown tickets your AI coding agents leave behind. Reads files directly — Claude, Codex, Gemini, Copilot, local models all push to the same folder.

Public API + agent skill documentation for RiskState — pre-trade risk governance for BTC and ETH crypto trading agents. Spot, perps, DeFi aware.

Reference architectures combining open agent frameworks, market intelligence, and execution rails with RiskState pre-trade risk governance.

MCP server for RiskState — pre-trade risk governance API for BTC and ETH crypto trading agents. Spot, perps, and DeFi aware.

Self-guided workshop for everyone at GitLab, regardless of technical background, who is interested in using agents to produce high-quality work.

GitLab CI/CD component that runs trustabl — static reliability/safety analyzer for AI agent SDKs (Claude, OpenAI, Google ADK, MCP). Gates pipelines on risk + severity.

https://github.com/trustabl/trustabl

Truitr is an AI-powered social network where users and autonomous agents can create content, interact through posts, comments, reactions and direct messages, and communicate across multiple social network instances through interoperability APIs. Built with Django, PostgreSQL, Azure AI, WebSockets and Docker.

Virtual Company Kernel is a public framework for running software delivery with AI agents under clear governance, human approvals, reusable workflows, and auditable execution artifacts.

AI-powered security orchestration for GitLab CI/CD. Automated vulnerability patching, threat modeling, and compliance scoring with GPT-4 and Claude AI agents.

A Coinbase x402 micropayment-gated API proxy for The Undesirables TCG Oracle. Allows AI agents to semantically discover and autonomously pay for physical trading card grading and Monte Carlo market intelligence using USDC on the Base network. Fully integrated with the CDP Bazaar Discovery layer.

Persistent and containerized development environment for Gemini CLI (https://geminicli.com/). This project provides a secure, automated workspace featuring built-in skill management (context-manager for memory, gema-manager for agent orchestration) and full support for MCP (Model Context Protocol). Designed for seamless local development, robust state persistence, and modular agent integration.

Use Case Study: Create Agentic IA With Strans Agents SDK and AgentCore Runtime on AWS

MCP-native browser automation server — 62 composite tools, 153 CLI subcommands, zero config. Give AI agents full browser control via the Model Context Protocol. Navigate, interact, extract text/HTML/screenshots, debug network, manage sessions, visual diffing, React inspection, auth vault, HAR recording. Works with Claude, Copilot, VS Code, Cursor. Built on Playwright/Chromium. The most comprehensive browser MCP server available.

Coordinated Agent Swarm Testing — open-source framework that deploys 100 Claude AI agents in five specialized squads (Security, Logic, Robustness, Performance, Data Integrity) to systematically analyze codebases for bugs, security issues, and code quality problems. A research-backed substitute for human beta testing. Empirical results from the published paper: 100 agents analyzed a 12,500-line React/TypeScript application in 63 minutes, identified 687 unique findings, at a total cost of $7.97 — compared to $2,000-$20,000 for equivalent human beta testing. Available via npm (cast-swarm) and as an MCP server. Apache 2.0 licensed. Published preprint available.

Governance framework for multi-agent AI systems in software development

A modular, secure, and production-ready MCP (Model Context Protocol) server designed for AI agent integration. Features dynamic tool registration, automated documentation indexing (RAG), and hardened security with TLS/SSL support.

Portable skill for adding a Claude Code lane to OpenAI Symphony + Linear workflows

The Governed Agentic AI Operating System — Rust + Tauri 2.0 | 65 crates, 658 commands, 84 pages, 5,029 tests, 10/10 OWASP