OpenAI Agents SDK for tools and handoffs

How it works

The SDK's runner coordinates agent execution, including tool calls and handoffs to other agents. Tracing makes those transitions visible when inspecting behavior across a run.

What you need

Start with a runner, a model and one documented tool. Review which telemetry configuration is active before sending test data you expect to remain local.

What to try first

Connect a read-only tool to one agent and build a second agent for one well-defined handoff. Test a request near the boundary to see who handles it and what context crosses over.

Use a handoff test with a synthetic account identifier. Inspect whether the destination agent receives it, and verify that the receiving tool checks permissions independently.

Where to look for failures

A handoff changes control flow, not user permissions. Limit the data and capabilities given to each agent, and review trace handling before testing with sensitive prompts.

Related Franklin pages

Compare who owns each step and which tools are available at a handoff. A shorter trace with the same validated result can be easier to operate than a larger team of agents.

• Build agents with tools and knowledge sources, then organize them into teams or workflows. Agno for agents, teams and workflow services
• Assign model-backed roles and tasks within a coordinated workflow. CrewAI for role-based agent workflows

Browse Franklin for concrete examples of tools and services exposed through MCP. Franklin’s MCP directory

Project documentation

• Current project README
• README at the reviewed revision