Connect agents to Microsoft Foundry Toolboxes via a Streamable HTTP MCP endpoint.
npx skills add https://github.com/aiappsgbb/awesome-gbb --skill foundry-toolbox
Please help me install this Agent Skill. Skill: foundry-toolbox Source: https://github.com/aiappsgbb/awesome-gbb/tree/main/skills/foundry-toolbox Command: npx skills add https://github.com/aiappsgbb/awesome-gbb --skill foundry-toolbox
The foundry-toolbox Skill solves the challenge of wiring multiple Microsoft Foundry tools into an agent without duplicating credentials, authentication logic, and version management across each tool.
Use the foundry-toolbox Skill to connect a code-based agent to the hosted toolbox MCP endpoint and let the platform handle credentials and versioning for a multi-tool workflow.
High-intent search queries and answers about installing and using this skill.
You can connect a code-based agent to multiple Microsoft Foundry tools by using a single Streamable HTTP MCP endpoint that bundles web search, Azure AI Search, code interpreter, and OpenAPI tools together. This centralized approach eliminates the need to wire each tool individually.
Unauthorized toolbox calls typically occur when the mandatory preview opt-in header Foundry-Features: Toolboxes=V1Preview is missing or the AAD token scope is handled incorrectly. You must include this specific header and ensure correct token injection for successful authentication.
Yes, the hosted Microsoft Foundry toolbox MCP endpoint supports integration with code-based agent architectures including LangGraph, GitHub Copilot SDK, and MAF. The platform provides specific integration patterns and guardrails to prevent common failure modes for these frameworks.
Centralized credential injection is the process of managing authentication tokens and refresh logic in one place for a bundle of MCP tools. You need this when wiring multiple Microsoft Foundry tools into an agent to avoid duplicating authentication code across each individual tool.
No, the Microsoft Foundry toolbox MCP endpoint explicitly skips unsupported MCP operations like ping and prompts list. You must use streaming tool calls instead and adhere to these connector-specific constraints to avoid errors during agent workflows.
The best way to manage versioning is to use the default and version-pinned endpoints provided by the Foundry toolbox. This allows you to test new versions safely and promote them deliberately without breaking existing agent integrations.
syeda-hoorain-aliCreate and configure Streamable HTTP MCP server connections for OpenAI Agents SDK.
aiappsgbbCreate and manage Microsoft Foundry prompt agents with model instructions and tool wiring.
aiappsgbbDeploys Microsoft Foundry hosted agents using MAF Agent framework variants.
microsoftBuild and manage Foundry agents with the Microsoft Agent Framework .NET SDK.
