
Swap run() for stream() and watch every reasoning step, tool call, and result land in real time — without giving up Agentspan's durability.
Read more
Build an issue triage bot in Agentspan with Strategy.ROUTER, where a dedicated classifier agent routes each issue to the right specialist — cheaper...

Build a human-directed editorial workflow in Agentspan with Strategy.MANUAL, where you pick which agent runs next — writer, fact checker, or copy...

Build a multi-agent brainstorming session in Agentspan with Strategy.RANDOM, where agents are selected randomly each turn to generate diverse,...

Build a code review debate in Agentspan with Strategy.ROUND_ROBIN, where specialist agents take turns, respond to each other, and converge on a...

Build a support swarm in Agentspan using Strategy.SWARM so specialist agents can transfer work directly to each other.

A deep dive into a durable coding assistant REPL built with Agentspan, covering the agent loop, tools, event streaming, workflow message queues,...

A practical look at why durable execution, server-side state, and workflow observability matter when AI agents move from demos to production.
Use Agentspan `@tool` functions and `ToolContext.state` to compare SBOMs, apply review policy, and keep the full execution trail on the server.

Use Agentspan `api_tool()` to turn an existing OpenAPI spec into agent-callable operations, scope the API surface with `tool_names`, and inspect the...