14 KiB
name, description
| name | description |
|---|---|
| microsoft-docs | Query official Microsoft and Microsoft-adjacent documentation to understand concepts, find tutorials, and get code examples. Covers Azure, .NET, Agent Framework, Semantic Kernel, DevUI, Aspire, VS Code, GitHub, Microsoft 365, Windows, and Power Platform using Microsoft Learn MCP and Context7 for comprehensive lookup across learn.microsoft.com, code.visualstudio.com, docs.github.com, aspire.dev, and GitHub source repos. |
Microsoft Docs
A unified documentation research skill for the full Microsoft technology ecosystem — including technologies like VS Code, GitHub, and Agent Framework repos that live outside or ahead of learn.microsoft.com.
Tools
Microsoft Learn MCP Server
| Tool | Use For |
|---|---|
microsoft_docs_search |
Find documentation on learn.microsoft.com — concepts, guides, tutorials, configuration |
microsoft_code_sample_search |
Find working code snippets from official Microsoft Learn docs. Especially strong for Agent Framework, Semantic Kernel, DevUI, and Azure SDK examples. Pass language parameter (python, csharp) for best results. |
microsoft_docs_fetch |
Get full page content from a specific Microsoft docs URL (when search excerpts aren't enough) |
Context7 (broader coverage)
Context7 indexes both websites and GitHub repos, giving access to documentation that the Microsoft Learn MCP server cannot reach — including VS Code docs, GitHub docs, Aspire.dev, GitHub CLI, and README-level content from source repos that is often ahead of published docs.
| Tool | Use For |
|---|---|
mcp_context7_resolve-library-id |
Find the Context7 library ID for a technology (one-time per session) |
mcp_context7_query-docs |
Query docs with a specific library ID — returns code snippets, explanations, and source links in a single call |
When GitHub Repo Sources (Context7) Beat Published Docs
Context7 indexes GitHub repo README files, which often contain API-level detail, CLI references, and advanced config that hasn't been published to learn.microsoft.com yet. This is especially true for:
- DevUI — the repo README (
/microsoft/agent-framework) has complete API reference (endpoints, parameters, response schemas), OpenAI proxy config, and auth setup that the Learn page summarizes - Agent Framework .NET — the repo has
Microsoft.Agents.AI.DevUIintegration code for ASP.NET Core before it appears on Learn - Extension samples —
/microsoft/vscode-extension-sampleshas working projects not always linked from the VS Code API docs
Rule of thumb: query the website source for tutorials and concepts, then the repo source for API-level detail and the latest features.
When to Use Which Tool
| Scenario | Best Tool | Why |
|---|---|---|
| Azure services, .NET, M365, Power Platform | microsoft_docs_search |
Direct access to learn.microsoft.com |
| Code samples for any Learn topic | microsoft_code_sample_search |
Returns up to 20 working code snippets with links. Filter by language. |
| Agent Framework — tutorials, concepts | microsoft_docs_search + Context7 Learn |
Polished guides on learn.microsoft.com |
| Agent Framework — DevUI setup, API, CLI | Context7 /microsoft/agent-framework |
GitHub repo has detailed DevUI README with full API reference |
| Agent Framework — DevUI tracing, directory discovery | Context7 /websites/learn_microsoft_en-us_agent-framework |
Learn website has step-by-step guides |
| Semantic Kernel — agents, plugins, orchestration | Context7 /websites/learn_microsoft_en-us_semantic-kernel + microsoft_code_sample_search |
Learn has the best SK docs; code sample search excels here |
| Semantic Kernel — source code, ADRs, design docs | Context7 /microsoft/semantic-kernel |
GitHub repo has samples and architectural decisions |
| VS Code features, settings, shortcuts | Context7 /websites/code_visualstudio |
VS Code docs live on code.visualstudio.com, not Learn |
| VS Code extension development / API | Context7 /websites/code_visualstudio_api |
Extension API docs are separate from user docs |
| GitHub Actions, repos, API, features | Context7 /websites/github_en |
GitHub docs live on docs.github.com |
| GitHub CLI (gh) | Context7 /websites/cli_github |
CLI reference lives on cli.github.com |
| .NET Aspire (AppHost, integrations, CLI) | Context7 /microsoft/aspire.dev |
Aspire docs live on aspire.dev, not Learn |
| Aspire community integrations (Go, Java) | Context7 /communitytoolkit/aspire |
Community Toolkit is a separate repo |
| Need full tutorial with all steps | microsoft_docs_fetch |
Gets complete page content |
Context7 Library Reference
Call mcp_context7_resolve-library-id once per session to confirm IDs. Below are tested, quality-ranked sources.
Agent Frameworks
Microsoft Agent Framework
| Library ID | Snippets | Score | Coverage |
|---|---|---|---|
/websites/learn_microsoft_en-us_agent-framework |
2,282 | 81.2 | Primary for tutorials — DevUI guides, directory discovery, tracing, AG-UI integration, workflow orchestration |
/microsoft/agent-framework |
1,177 | — | Primary for API detail — DevUI README (full REST API, CLI flags, auth, OpenAI proxy), .NET DevUI integration, samples |
The Learn website source has polished tutorials and conceptual docs. The GitHub repo source has README-level detail that is often ahead of published docs — particularly the DevUI REST API reference, advanced CLI options, and .NET Microsoft.Agents.AI.DevUI package setup.
Use both for comprehensive Agent Framework coverage: query the website source for "how to" guides, then the repo source for API-level specifics.
Researching DevUI
DevUI docs are split across Learn and the GitHub repo. The best research strategy:
| What You Need | Where to Search |
|---|---|
| Setup tutorial, directory structure | Context7 /websites/learn_microsoft_en-us_agent-framework — query "DevUI directory discovery" or "DevUI tracing" |
| CLI flags, REST API endpoints, auth config | Context7 /microsoft/agent-framework — query "DevUI CLI options" or "DevUI POST /v1/responses" |
| Working Python code snippets | microsoft_code_sample_search with language: "python" — query "DevUI agent framework serve" |
| Working C# integration code | microsoft_code_sample_search with language: "csharp" — query "DevUI AddDevUI MapDevUI" |
| Full setup guide in one page | microsoft_docs_fetch with URL https://learn.microsoft.com/en-us/agent-framework/user-guide/devui/ |
The GitHub repo source is particularly valuable for DevUI because it contains API-level detail (endpoint schemas, proxy config, auth tokens) that the Learn page summarizes.
Azure & .NET (learn.microsoft.com)
| Library ID | Snippets | Score | Coverage |
|---|---|---|---|
/websites/learn_microsoft_en-us_azure |
164,912 | 73.8 | All Azure services — compute, storage, networking, AI, databases |
/microsoftdocs/azure-docs |
61,791 | 76.7 | Azure docs GitHub repo — higher quality score |
/websites/learn_microsoft_en-us |
158,360 | 45.4 | All of Microsoft Learn — broadest but lower precision |
/websites/learn_microsoft_en-us_dotnet_azure |
1,156 | — | Azure SDK for .NET developers specifically |
.NET Aspire
| Library ID | Snippets | Score | Coverage |
|---|---|---|---|
/microsoft/aspire.dev |
1,865 | 73.0 | Primary — official docs site repo, best quality, full Aspire 13+ coverage |
/websites/aspire_dev |
31,845 | 71.6 | Website crawl — massive but includes multilingual duplicates |
/dotnet/aspire |
1,185 | 71.5 | Runtime source — API internals, playground examples |
/communitytoolkit/aspire |
311 | 64.2 | Community integrations — Golang, Java, Node.js, Vite, Ollama |
/dotnet/docs-aspire |
482 | 71.4 | Legacy — being superseded by aspire.dev, avoid for tutorials |
Visual Studio Code
| Library ID | Snippets | Score | Coverage |
|---|---|---|---|
/websites/code_visualstudio |
6,288 | 80.4 | Primary — user docs, settings, features, debugging, remote dev |
/websites/code_visualstudio_api |
1,681 | 65.4 | Extension API — webviews, TreeViews, commands, contribution points |
/microsoft/vscode-docs |
8,289 | 87.9 | Docs repo — highest score, markdown source files |
/microsoft/vscode-extension-samples |
320 | 82.4 | Extension samples — working code examples |
GitHub
| Library ID | Snippets | Score | Coverage |
|---|---|---|---|
/websites/github_en |
43,828 | 66.2 | Primary — Actions, API, repos, security, admin, Copilot |
/github/docs |
24,544 | 73.7 | Docs repo — markdown source, higher quality score |
/websites/cli_github |
386 | 83.2 | GitHub CLI reference — gh commands, flags, examples |
/websites/cli_github_manual |
478 | 44.0 | CLI manual — alternative, lower score |
Query Effectiveness
Good queries are specific:
# ❌ Too broad
"Azure Functions"
"VS Code extensions"
"agent framework"
# ✅ Specific
"Azure Functions Python v2 programming model"
"Cosmos DB partition key design best practices"
"VS Code webview API onDidReceiveMessage postMessage"
"GitHub Actions workflow_dispatch inputs matrix strategy"
"Aspire AddUvicornApp Python FastAPI integration"
"DevUI serve agents tracing OpenTelemetry directory discovery"
"ChatCompletionAgent plugin tool calling multi-agent orchestration"
"Agent Framework workflow conditional edges branching handoff"
Include context:
- Version when relevant (
.NET 8,Aspire 13,VS Code 1.96) - Task intent (
quickstart,tutorial,overview,limits,API reference) - Platform for multi-platform docs (
Linux,Windows,macOS) - Language for polyglot docs (
Python,TypeScript,C#)
Agent Framework Query Examples
# DevUI — setup and running
microsoft_code_sample_search(language="python", query="DevUI agent framework serve agents tracing")
context7_query("/microsoft/agent-framework", "DevUI CLI options port host headless authentication")
context7_query("/websites/learn_microsoft_en-us_agent-framework", "DevUI directory discovery agent workflow __init__.py")
# DevUI — tracing and observability
context7_query("/websites/learn_microsoft_en-us_agent-framework", "DevUI tracing OpenTelemetry debug panel span hierarchy")
microsoft_docs_search("configure tracing agent framework DevUI OTLP_ENDPOINT Jaeger")
# DevUI — API and OpenAI SDK
context7_query("/microsoft/agent-framework", "DevUI POST /v1/responses OpenAI SDK entity_id streaming")
context7_query("/microsoft/agent-framework", "DevUI OpenAI proxy X-Proxy-Backend authentication")
# DevUI — .NET integration
microsoft_code_sample_search(language="csharp", query="DevUI AddDevUI MapDevUI ASP.NET Core agent")
context7_query("/microsoft/agent-framework", "Microsoft.Agents.AI.DevUI AddDevUI MapDevUI ASP.NET Core")
# Semantic Kernel agents
microsoft_code_sample_search(language="python", query="ChatCompletionAgent plugin AgentGroupChat")
context7_query("/websites/learn_microsoft_en-us_semantic-kernel", "ChatCompletionAgent plugin tool calling setup")
context7_query("/websites/learn_microsoft_en-us_semantic-kernel_frameworks_agent", "AgentGroupChat selection strategy termination")
# Agent Framework workflows
context7_query("/websites/learn_microsoft_en-us_agent-framework", "workflow orchestration multi-agent handoff conditional edge")
microsoft_code_sample_search(language="csharp", query="agent framework workflow WorkflowBuilder AddEdge")
# Cross-framework tracing
microsoft_docs_search("configure tracing AI agent frameworks Foundry Semantic Kernel Agent Framework")
Workflow
- Identify the technology — is it Azure/Learn content, Agent Framework, Semantic Kernel, DevUI, VS Code, GitHub, or Aspire?
- Pick the right tool — use the decision table above to choose between Microsoft Learn MCP and Context7
- For Agent Frameworks — use
microsoft_code_sample_searchfor working code, then Context7 for deeper API docs:- DevUI how-to: Context7 Learn website → code sample search for working examples → Context7 repo for API detail
- Semantic Kernel: Context7 Learn website + code sample search → Context7 repo for advanced samples
- Workflow orchestration: Microsoft Learn search → Context7 Learn website for patterns
- For Context7 — resolve the library ID first (if not cached), then query with a specific question
- For Microsoft Learn — search first, then fetch full pages when you need complete tutorials or all config options
- Combine tools — for cross-cutting questions (e.g., "deploy Aspire to Azure", "trace Agent Framework to Azure Monitor"), use multiple sources
When to Fetch Full Page (Microsoft Learn MCP)
Use microsoft_docs_fetch after microsoft_docs_search when:
- Tutorials — need complete step-by-step instructions
- Configuration guides — need all options listed
- Deep dives — user wants comprehensive coverage
- Search excerpt is cut off — full context needed
- Agent Framework DevUI docs —
https://learn.microsoft.com/en-us/agent-framework/user-guide/devui/for complete setup guide
Why Use This
- Accuracy — live docs and indexed sources, not stale training data
- Completeness — full tutorials, not fragments
- Breadth — covers the entire Microsoft ecosystem including VS Code, GitHub, Agent Framework repos, and Aspire that live outside learn.microsoft.com
- Depth — GitHub repo sources provide API-level detail ahead of published docs
- Authority — official Microsoft and first-party documentation