Guild.ai

A manual run is directly triggered by a user for testing purposes and is available on the Free plan. An automation is an event-driven execution triggered by system events such as a GitHub pull request, a Jira issue creation, or a PagerDuty alert. Persistent automations are available from the Individual plan upward.

No. Guild.ai is fully model-agnostic. Teams can bring their own API keys and use OpenAI (GPT-4o, o1, o3), Anthropic (Claude 4 Opus, Sonnet, Haiku), or Google (Gemini) models. Guild handles governance, access control, and observability regardless of which model provider is used.

Guild uses a credentials-as-code model where permissions are declared at the agent level in YAML. Each agent is assigned scoped credentials that specify exactly which API endpoints it can access—for example, only `pulls:read` on a specific GitHub org—preventing agents from accessing systems beyond their defined scope.

Yes. The Agent Hub allows teams to publish agents as public or private. Any agent can be forked and customized for a specific use case, then re-published. This makes it practical for organizations to build a shared internal library of production-tested agents rather than rebuilding from scratch each time.

The Guild Insights Dashboard provides a single view of how AI agents consume resources across an organization. It surfaces per-agent token usage, automation counts, cost attribution, and spend trends—giving engineering and finance teams the visibility needed to audit agent activity and prevent cost overruns at scale.