microsoft/agent365-runbook

microsoft/agent365-runbook

Description: Step-by-step runbooks for bringing your own agent into Microsoft Agent 365: identity, observability, Microsoft 365 data access, and messaging.

Language: TypeScript

Stars: 2

Forks: 0

Open issues: 9

Created: 2026-09-23T14:34:21Z

Pushed: 2026-10-02T06:40:53Z

Default branch: main

Fork: no

Archived: no

README:

Agent 365 Runbooks

Step-by-step runbooks for bringing your own agent into Microsoft Agent 365: identity, observability, Microsoft 365 data access, and messaging.

We can onboard our own working agent or use an optional educational sample. All four scenarios have a skill-led runbook and a manual runbook, with .NET, Python and Node.js examples. The identity path depends on who acts; the choice of a coding assistant does not change it.

Choose how to work

Use 3.Runbook.md to work with a coding assistant, or 3.Runbook-Manual.md to run the CLI and make the source edits ourselves. Both routes cover registration and observability. Work IQ requires separate permissions and runtime tool integration where supported; the S2S scenario does not add it.

With a coding assistant

The skill-led runbooks use the Agent 365 Skills, installed into our chosen assistant:

| Your coding assistant | Install with | | --- | --- | | Claude Code (app, web, CLI) | /plugin marketplace add https://github.com/microsoft/agent365-skills then /plugin install agent365@agent365-skills | | GitHub Copilot CLI, VS Code agent mode | gh skill add microsoft/agent365-skills | | Cursor, Windsurf, Codex CLI, Gemini CLI, and other agentskills.io-compatible tools | node /path/to/agent365-skills/scripts/install.js from your project, which installs into .agents/skills/ |

Full instructions, including how to verify the install: [Installing the Skills](00-overview/Installing-the-Skills.md).

| Skill | What it does | | --- | --- | | a365-setup | Installs the Agent 365 CLI, validates Azure prerequisites, detects your stack, routes you to the right path | | make-a365-agent | Registers a Blueprint for agents needing observability or catalog visibility, without the messaging layer | | instrument-observability | Wires OpenTelemetry and the Agent 365 tracing exporter | | add-workiq-tools | Adds selected Work IQ MCP servers and supported runtime wiring; delegated scenarios only | | make-ai-teammate | Adds Messaging and Notifications so the agent can receive Teams messages, email and @mentions | | test-local | Launches an AI Teammate with AgentsPlayground; S2S uses HTTP requests instead |

The skill-led steps also explain the commands and code behind the automation:

> What you type → what the skill does → the CLI commands and SDK code behind it → how to verify it worked

Review generated changes against the scenario's identity and telemetry requirements before deployment. The [known skill gaps](03-references/Known-Skill-Gaps.md) describe integration details to check.

Without a coding assistant

The manual runbooks provide the portal steps, CLI commands and source edits without requiring skills or a coding assistant. Choose the manual link for our scenario below; we don't need to read the skill-led guide first.

For registration, authentication or missing activity, use the [troubleshooting guide](03-references/Troubleshooting.md). It covers all four scenarios and both authoring routes, with checks for the running process, token flow and exported spans.

Scenarios

| Scenario | Identity and hosting | Skill-led runbook | Manual runbook | | --- | --- | --- | --- | | [Web app: User OBO](01-scenarios/Web-App-Agent-User-OBO/1.Overview.md) | Signed-in user's delegated permissions; web app | [Skill-led](01-scenarios/Web-App-Agent-User-OBO/3.Runbook.md) | [Manual](01-scenarios/Web-App-Agent-User-OBO/3.Runbook-Manual.md) | | [Teams: Custom engine OBO](01-scenarios/Teams-Agent-Custom-Engine-OBO/1.Overview.md) | Teams user's delegated permissions; Azure Bot channel | [Skill-led](01-scenarios/Teams-Agent-Custom-Engine-OBO/3.Runbook.md) | [Manual](01-scenarios/Teams-Agent-Custom-Engine-OBO/3.Runbook-Manual.md) | | [AI Teammate / Autopilot](01-scenarios/AI-Teammate-Agent-Identity/1.Overview.md) | Agent's own user account; Microsoft 365 presence | [Skill-led](01-scenarios/AI-Teammate-Agent-Identity/3.Runbook.md) | [Manual](01-scenarios/AI-Teammate-Agent-Identity/3.Runbook-Manual.md) | | [Service-to-service agent](01-scenarios/Service-to-Service-Agent/1.Overview.md) | App-only permissions, no signed-in user; webhook or background work | [Skill-led](01-scenarios/Service-to-Service-Agent/3.Runbook.md) | [Manual](01-scenarios/Service-to-Service-Agent/3.Runbook-Manual.md) |

Not sure which applies to you? Start with [Choosing Your Onboarding Path](02-patterns/Choosing-Your-Onboarding-Path.md), which walks the decision and lists what each path costs you.

Repository structure

| Path | Contents | | --- | --- | | [00-overview/](00-overview/) | Onboarding concepts and the choice between skill-led and manual authoring | | [01-scenarios/](01-scenarios/) | Four scenarios, each with both runbook paths and optional sample code | | [02-patterns/](02-patterns/) | Cross-cutting guidance that applies to more than one scenario | | [03-references/](03-references/) | Troubleshooting, known skill gaps, and environment gotchas |

Each scenario folder follows the same layout:

/
├── 0.Resources/
│ ├── Images/
│ └── Starting-point/ ← optional educational agents without Agent 365 instrumentation
├── 1.Overview.md ← what this scenario is and who it is for
├── 2.Architecture.md ← how the pieces fit, and the token flow
├── 3.Runbook.md ← skill-led onboarding
├── 3.Runbook-Manual.md ← manual registration and instrumentation
└── 4.Sample-prompts.md ← chat prompts or S2S HTTP requests for the sample

Both runbooks use the same scenario identity model. The sample requests exercise the behavior the chosen implementation actually supports; a stub notification is not a real Teams message.

Prerequisites

Start with a working agent or an optional sample, then follow the prerequisites in the chosen runbook:

• A Microsoft 365 tenant with Agent 365 enabled, appropriate licensing and permission to register the agent and complete administrator consent.
• Access to the resources our agent uses. Azure model resources are needed only when using that provider; the S2S samples also have a...