Visibility
Know what agents are doing
See exactly what every agent did at runtime—from each tool call to every token spent, and more. All downloadable for distribution across your organization
Monitor every agent at runtime
Catch issues, maintain complete audit trails, identify waste, and improve reliability and trust across your agent fleet.
Full session traces for every run
Every model, tool call, and decision is captured and available in real time.
Credentials
Maintain accurate and complete audit trails
When a SOC 2 auditor requests evidence, retrieve the complete trace directly from the CLI.
Approvals
Control spend with Circuit Breaker
Track token usage by workspace, agent, user, provider, and model. Set limits and stop unexpected spending before it becomes an overrun.
Programmatically evaluate and control agent performance
Automatically deploy evals that score agent outputs against defined criteria, identify where performance is falling short, and stop agents that fail critical checks.
Frequently asked questions
AI agent observability is the ability to see exactly what an agent did at runtime: which model calls it made, which tools it invoked, what inputs it received, what outputs it produced, and what decisions it took along the way. Without it, debugging an agent failure or auditing an agent action is guesswork.
Full session traces: every model call, every tool invocation, every input and output, every credential attached, every decision taken. Traces are captured in real time and available for search, export, or handoff to a SOC 2 auditor. Nothing about the run is opaque.
Guild lets you define eval criteria (behavioral, structural, latency, cost) and automatically scores every agent run against them. When an agent fails a critical check, the runtime can stop it or flag it for review. Evals turn observability from passive logging into active quality enforcement.
Yes. Any agent that runs on the Guild runtime, regardless of the framework it was built in (LangChain, CrewAI, custom code, or others), gets the same session traces, audit trails, and eval scoring. Observability is applied at the runtime level, not per framework.
Every agent action produces an immutable audit trail with timestamps, acting user, credentials used, tools called, and outputs generated. When an auditor asks for evidence of what an agent did, retrieve the complete trace directly from the CLI or dashboard. Guild is SOC 2 Type 1 compliant, ISO 27001 certified, and GDPR compliant.
Run. Control.
Transform with Guild.
See Guild in action.