Changelog

What’s New

Triggers for Agents Triggers can now start an agent, not just a task. Fire an agent from an incoming webhook, or run it automatically on a recurring schedule with cron and time zone support. Webhook-triggered agent runs are accepted immediately and processed in the background, so long-running work never blocks the caller. See Triggers and runs.

Sequential Execution Mode Agents now run in one of two modes. In agentic mode the model decides which tools to call and in what order; in sequential mode the agent runs a fixed pipeline where each step feeds the next. Choose the mode that fits the job when you build an agent.

Direct MCP Server Connections Attach an MCP server straight to an agent for direct, single-hop access to its tools, with no intermediate task required. A redesigned MCP servers area and marketplace make it easier to add, browse, and connect servers. See MCP.

Agent Observability and Reporting New token, credit, timing, and usage reports give you an aggregate view of how your agents perform. Run, session, and event data is now enriched with per-step timing and error detail, so you can trace a run without opening each underlying task. See Observability.

Improvements

• A new versioning interface for reviewing agent revisions and rolling back to an earlier one. See Versioning and revisions.
• Improved retrieval-augmented generation for tasks backed by a knowledge base.
• Filter the tool list when configuring an agent, making large tool sets easier to work with.
• Redesigned agent pages.
• Hardened handling of webhook credentials.