The Control Plane for AI Agents Is Now Open

We’re opening Guild to everyone today.

AI agents are already in production. They review code, triage issues, update systems, and automate workflows across the enterprise. As adoption accelerates, the problem changes. It is no longer whether agents are useful. It is whether you can control them once they start executing actions in your systems.

Most teams can’t.

Agents are built across different frameworks, connected to different tools, and deployed without a consistent governance model. There’s no reliable way to see what they’re doing, control what they can access, or reuse what already works. What starts as helpful automation quickly becomes difficult to manage at scale.

At this point, it becomes an infrastructure problem.

Now, for the first time, there is a control plane for AI agents. As agents move into production, this layer sits between models and enterprise systems, governing how autonomous software runs.

Guild is that layer. It’s what’s been missing - and this is a significant development for managing the complete agent lifecycle.  

With Guild, agents do not operate directly on your systems. Every execution runs through Guild, where identity is enforced, access is controlled, and every action is recorded and traceable. This is what makes agents safe to run in production.

Guild defines the control plane for AI agents.

Guild includes a governed runtime for executing agents, ready-to-run starter agents, a Managed Agent Center for internal reuse, and an Agent Hub for sharing capabilities across teams and organizations.

Running agents in production requires:

• A defined identity for every agent
• Centralized control over what systems each agent can access
• Visibility into what actually happened, not just outputs
• The ability to update and roll back safely
• A way to reuse what already works

Guild provides these capabilities in one place.

Most agents today are one-off tools built for a single workflow. That model doesn’t scale.

Production systems are multiplayer. They depend on shared infrastructure, governance, and reuse.

Guild turns agents into shared systems. Teams can publish, version, and reuse agents through the Managed Agent Center. Agent Hub extends that beyond the organization, making capabilities discoverable and reusable across environments.

What works in one place can work everywhere.

That’s how capabilities become shared infrastructure.

The Open Beta includes the Guild control plane, starter agents, Agent Hub, integrations, and developer tools for building or bringing your own agents.

You can start with agents like:

• Ticket Processing: triages Jira and GitHub issues and routes them to the right owner
• Code Review: reviews pull requests and posts feedback
• Slack Bot: turns Slack conversations into structured work
• Tech Support: answers questions using internal documentation

These agents are versioned and governed by default. You can deploy them, fork them, adapt them, or bring your own agents and run them through the same control plane.

Guild integrates with systems teams already use, GitHub, Bitbucket, Azure DevOps, Jira, Linear, Slack, Notion, Zendesk, TestRail, New Relic, Google Compute, Google Docs, and Google Logging. Agents can securely access these services through OAuth, with permissions enforced at the control plane level. All access is mediated through the control plane, so permissions are enforced consistently, and API keys are not exposed.

If a system isn’t supported out of the box, you can extend Guild by registering custom APIs as tools or building agents with the SDK and CLI.

Models will change. The control layer can’t.

Guild works across models, vendors, and frameworks. Teams can use OpenAI, Anthropic, Google, or other providers without changing how governance is applied. Every call flows through the control plane, giving teams visibility and control as the underlying intelligence evolves.

Security is built into the system.

With Guild, agents run in controlled environments. Credentials are scoped and mediated. Organizations are isolated. Every action is recorded with full context, so teams can monitor behavior, usage, and cost in real time. This is the baseline for running agents in production.

AI adoption is already happening. What’s missing is the infrastructure to manage it. With Guild, this layer now exists.

Guild is the control plane that brings structure, visibility, and governance to how agents are built, deployed, shared, and used.

If you’re building agents, you’re going to run into this problem. You can start using Guild today at guild.ai. Let us know what you think.