version: "1.0.0" name: microsoft-toolbox description: >- Help a learner use Microsoft Toolbox to define tools once and govern them centrally, then consume a toolbox from an agent through a single MCP endpoint. Use for Lesson 6 tasks: tool discovery, registration, governance, and versioning.

Microsoft Toolbox

Guidance for centralised tool governance (Lesson 6).

Guardrails (always apply)

• Terminology: Microsoft Toolbox (define/govern tools centrally), consumed by agents over an

MCP endpoint.

• The consume side works with the pinned agent-framework SDK. The Toolbox management API

(project.toolboxes.*) is preview and is not in the pinned SDK — call this out and avoid presenting it as GA.

• Endpoint via TOOLBOX_ENDPOINT in .env; auth via az login.

Concepts a learner should grasp

1. Why a toolbox: define a tool once, govern it centrally (access, versioning, approvals), and

let many agents consume it — instead of each agent re-implementing tools.

1. Consume via one MCP endpoint: an agent points at the toolbox's MCP endpoint and gets its

governed tools, rather than wiring each tool individually.

1. Governance & versioning: central control makes it safe to update or retire a tool without

breaking every agent.

Worked example in this repo

• lesson-6-toolbox/toolbox_agent.py is the consume sample: it reaches a toolbox through a

single MCP endpoint and uses its tools. Use it as the reference.

How to help

• Start from toolbox_agent.py; adapt the toolbox endpoint.
• If the learner asks to create/manage toolboxes in code, explain the management API is preview and

may require a newer SDK than the pinned one; prefer the portal/preview docs and keep the sample on the consume side.

Validate

• python -m py_compile lesson-6-toolbox/toolbox_agent.py.
• The agent should discover and call at least one governed tool via the toolbox endpoint.

References

• lesson-6-toolbox/README.md