Open Source AI Agents

AI agents are software systems that do more than answer prompts. They plan, call tools, read and write data, and carry out multi-step tasks on a user's behalf. The open source side of this field covers openly licensed models, frameworks, and interoperability standards that let developers build, inspect, and connect agents without being locked into a single vendor.

Why it matters now

Agents are moving from demos into real products, and the plumbing behind them is being standardized in public. Protocols such as the Model Context Protocol (MCP) aim to give agents a common way to reach apps, data sources, and services. That shared layer speeds up adoption, but it also widens the attack surface. Every new connector, server, or marketplace listing is another place where permissions, data access, and trust can go wrong.

Large platforms and smaller projects alike are competing to shape these standards. Competing proposals for how personal agents should communicate are a sign the ecosystem has not yet settled on a single approach. Openly released models built for agent decision-making also face a practical test: whether developers actually adopt them.

What you'll find here

• Protocols and standards: how agents connect to tools and to each other, and where specifications diverge
• Open models and frameworks: releases designed for reasoning, tool use, and orchestration
• Security and incidents: vulnerabilities, risky integrations, and lessons from real-world compromises
• Platform strategy: how major tech companies are weaving agents into messaging, productivity, and business software
• Developer tooling: marketplaces, low-code builders, and workflow automation

Follow this page to track how open agent technology is built and secured, and who ends up shaping it.