An Agent Builder Run is a complete invocation of an Agent Builder agent, initiated by the receipt of a message or event, whether triggered automatically or sent manually. An Agent Builder Run includes any message delivered to the agent through configured triggers (including third-party integrations or other connected services) as well as messages sent directly through the Agent Builder user interface (for example, via the chat interface). An Agent Builder Run is counted if the invocation succeeds or is interrupted. Resuming execution after a human-in-the-loop interruption constitutes a separate Agent Builder Run. An execution failure does not count as an Agent Builder Run.
Are model costs included with Agent Builder pricing?
No. Model usage is billed separately by your model provider. Agent Builder lets you choose the model you want (for example, Anthropic, OpenAI, Gemini, or another provider) and connect it using your provider API key. For more information on configuring models, see the Agent Builder setup page.
Are agent tools included with Agent Builder pricing?
Agent Builder includes built-in tools and supports connecting third-party tools via a remote MCP server. Third-party tools typically require you to authenticate with your account, and any fees or usage charges are billed by the third-party provider. For more information on setting up tool keys, see the Agent Builder setup page.
Are Agent Builder runs automatically traced in LangSmith, and do they count toward billing?
Yes. Every Agent Builder run is automatically traced in LangSmith. These traces are attributed to the user’s account and count toward usage-based billing under the customer’s LangSmith plan.
