msgraph

Up-to-date Microsoft Graph API knowledge for AI agents. Search 27,700+ Graph APIs, endpoint docs, resource schemas, and community samples — all locally, no network calls. Use when the agent needs to find, understand, or call Microsoft Graph endpoints.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "msgraph" with this command: npx skills add merill/msgraph/merill-msgraph-msgraph

Microsoft Graph Agent Skill

Search, look up, and call any of the 27,700+ Microsoft Graph APIs — all locally, no network calls needed. Use the three search commands to find the right endpoint, check permissions and parameters, then optionally execute calls directly or hand off to a Graph MCP server.

What's Included

The Microsoft Graph API has 27,700+ endpoints updated weekly — well past LLM training cutoffs. This skill bundles the complete API surface as local indexes that you search instantly with no network calls.

IndexCountWhat it contains
OpenAPI endpoints27,700+Path, method, summary, description, permission scopes
Endpoint docs6,200+Permissions (delegated/app), query parameters, required headers, default vs $select-only properties
Resource schemas4,200+All properties with types, supported $filter operators, default/select-only flags
Community samplesGrowingHand-verified queries mapping natural-language tasks to exact API calls

How to Run

The msgraph CLI is bundled with this skill. Run all commands through the launcher script in this skill's directory:

  • macOS / Linux: bash <path-to-this-skill>/scripts/run.sh <command> [args...]
  • Windows: powershell <path-to-this-skill>/scripts/run.ps1 <command> [args...]

For example, to search for mail-related APIs on macOS:

bash /home/user/.opencode/skills/msgraph/scripts/run.sh openapi-search --query "send mail"

In all examples below, msgraph is shorthand for the full launcher invocation.

Finding the Right API

This is the primary purpose of the skill. Follow this progressive lookup strategy — each level adds detail:

  1. Your own knowledge — try first for well-known endpoints (/me, /users, /groups).
  2. sample-search — curated, hand-verified samples. Highest quality. Use for common tasks and multi-step workflows.
  3. api-docs-search — per-endpoint permissions, supported query parameters, required headers, default vs $select-only properties, and resource property details with filter operators.
  4. openapi-search — full catalog of 27,700 Graph APIs. Use when you cannot find the endpoint any other way.
  5. Reference files — concept docs on query parameters, advanced queries, paging, batching, throttling, errors, and best practices. Read only when you need specific guidance.

This order is guidance — adapt based on the task. For example, jump straight to api-docs-search if you already know the endpoint but need its permissions.

sample-search

Search curated community samples that map natural-language tasks to exact Microsoft Graph API queries:

msgraph sample-search --query "conditional access policies"
msgraph sample-search --product entra
msgraph sample-search --query "managed devices" --product intune
FlagDescription
--queryFree-text search (searches intent and query fields)
--productFilter by product: entra, intune, exchange, teams, sharepoint, security, general
--limitMax results (default 10)

At least one of --query or --product is required. Results include multi-step workflows.

api-docs-search

Look up detailed documentation for a specific endpoint or resource type:

msgraph api-docs-search --endpoint /users --method GET
msgraph api-docs-search --resource user
msgraph api-docs-search --query "ConsistencyLevel"
FlagDescription
--endpointSearch by endpoint path (e.g. /users, /me/messages)
--resourceSearch by resource type name (e.g. user, group, message)
--methodFilter by HTTP method: GET, POST, PUT, PATCH
--queryFree-text search across all fields
--limitMax results (default 10)

At least one of --endpoint, --resource, or --query is required.

Endpoint results include: required permissions (delegated work/school, delegated personal, application), supported OData query parameters, required headers, default properties, and endpoint-specific notes.

Resource results include: all properties with types, supported $filter operators (eq, ne, startsWith, etc.), and whether each property is returned by default or requires $select.

openapi-search

Search the full OpenAPI catalog of 27,700 Microsoft Graph APIs:

msgraph openapi-search --query "send mail"
msgraph openapi-search --resource messages --method GET
FlagDescription
--queryFree-text search (searches path, summary, description)
--resourceFilter by resource name (e.g. users, groups, messages)
--methodFilter by HTTP method
--limitMax results (default 20)

At least one of --query, --resource, or --method is required.

Using with MCP Servers

If the agent has access to a Microsoft Graph MCP server (such as lokka.dev or any other Microsoft Graph MCP server), use the search tools above to find the right endpoint, permissions, and request syntax, then use the information with the MCP server for execution.

In this mode, no authentication through this skill is needed. The skill acts purely as a knowledge layer — the MCP server handles authentication and API execution.

Direct Microsoft Graph API Execution

When no Graph MCP server is available, this skill can authenticate to Microsoft 365 and execute Microsoft Graph API calls directly.

Authentication

The tool supports delegated (user) and app-only (application) authentication, auto-detected from environment variables.

Quick start:

msgraph auth status          # check if signed in
msgraph auth signin          # sign in (opens browser) - recommended
msgraph auth signin --device-code  # sign in via device code (headless)
msgraph auth signout         # clear the session
  • Delegated auth (default): Interactive browser sign-in, with device code fallback for headless environments. Supports incremental consent — on 403, the tool re-authenticates with required scopes and retries automatically.
  • App-only auth: Auto-detected when MSGRAPH_CLIENT_SECRET, MSGRAPH_CLIENT_CERTIFICATE_PATH, MSGRAPH_FEDERATED_TOKEN_FILE, or MSGRAPH_AUTH_METHOD=managed-identity is set. Requires MSGRAPH_TENANT_ID.

For detailed authentication configuration including certificates, managed identity, workload identity federation, and all environment variables, see references/docs/authentication.md.

Making Graph API Calls

IMPORTANT: Run msgraph auth status before the first graph-call in a session to verify authentication.

msgraph graph-call <METHOD> <URL> [flags]

Read Operations

msgraph graph-call GET /me
msgraph graph-call GET /users --select "displayName,mail" --top 10
msgraph graph-call GET /me/messages --filter "isRead eq false" --top 5 --select "subject,from,receivedDateTime"
msgraph graph-call GET /users --filter "startsWith(displayName,'John')"

Write Operations

IMPORTANT: YOU MUST ask the user for confirmation before any write operation. Write operations require the --allow-writes flag.

msgraph graph-call POST /me/sendMail --body '{"message":{"subject":"Hello","body":{"content":"Hi"},"toRecipients":[{"emailAddress":{"address":"user@example.com"}}]}}' --allow-writes
msgraph graph-call PATCH /me --body '{"jobTitle":"Engineer"}' --allow-writes

DELETE is always blocked regardless of flags.

graph-call Flags

FlagDescriptionExample
--selectOData $select--select "displayName,mail"
--filterOData $filter--filter "isRead eq false"
--topOData $top (limit results)--top 10
--expandOData $expand--expand "members"
--orderbyOData $orderby--orderby "displayName desc"
--api-versionv1.0 or beta (default: beta)--api-version v1.0
--scopesRequest additional permission scopes--scopes "Mail.Read"
--headersCustom HTTP headers--headers "ConsistencyLevel:eventual"
--bodyRequest body (JSON)--body '{"key":"value"}'
--outputjson (default) or raw--output raw
--allow-writesAllow POST/PUT/PATCH (requires user confirmation)

Critical Rules

Always (search and knowledge)

  1. Never guess or fabricate Microsoft Graph endpoints — always verify via search before calling. This skill exists because agents hallucinate endpoints; use it.
  2. Use the progressive lookup strategy — start with what you know, then sample-search, api-docs-search, openapi-search as needed.
  3. Use --select to reduce response size — only request fields you need.
  4. Use --top to limit results — avoid fetching thousands of records.
  5. ConsistencyLevel header is required for $count and $search on directory objects (users, groups, etc.). Use --headers "ConsistencyLevel:eventual".
  6. Default API version is beta — use --api-version v1.0 for production-stable endpoints.

When using direct execution (graph-call)

  1. Check auth status before the first graph-call in a session.
  2. GET is the default — no special flags needed.
  3. Write operations require --allow-writes — YOU MUST confirm with the user first.
  4. DELETE is always blocked — inform the user this is not supported.
  5. 403 triggers automatic re-auth — the tool requests additional scopes and retries (delegated auth only).
  6. All output is JSON — parse statusCode and body fields from the response.

Error Handling

StatusMeaningAction
401Token expiredRun msgraph auth signin again
403Insufficient permissionsTool auto-retries with incremental consent. If still fails, user needs admin consent.
404Resource not foundVerify the endpoint path
429Rate limitedWait for Retry-After duration, then retry

Environment Variables

VariableDescriptionDefault
MSGRAPH_CLIENT_IDCustom Entra ID app client IDMicrosoft Graph CLI Tools app
MSGRAPH_TENANT_IDTarget tenant ID (required for app-only)common
MSGRAPH_API_VERSIONDefault API versionbeta
MSGRAPH_INDEX_DB_PATHPath to OpenAPI index databaseAuto-detected
MSGRAPH_SAMPLES_DB_PATHPath to samples index databaseAuto-detected
MSGRAPH_API_DOCS_DB_PATHPath to API docs index databaseAuto-detected
MSGRAPH_NO_TOKEN_CACHEDisable persistent token cache (in-memory only)false

For the full list of authentication environment variables, see references/docs/authentication.md.

Compatibility

Search tools run fully offline with no network access required. Direct API execution requires network access to login.microsoftonline.com and graph.microsoft.com. A system browser is used for interactive auth; falls back to device code flow in headless environments.

Reference Files

Load these on demand when you need specific guidance. Do NOT load them preemptively.

FileWhen to ReadSize
references/REFERENCE.mdCommon resource paths, OData patterns, permission scopes~230 lines
references/docs/authentication.mdDetailed auth configuration: certificates, managed identity, workload identity, all env vars~200 lines
references/docs/query-parameters.mdOData $select, $filter, $expand, $top, $orderby, $search syntax and gotchas~300 lines
references/docs/advanced-queries.mdConsistencyLevel header, $count, $search, ne/not/endsWith on directory objects~190 lines
references/docs/paging.md@odata.nextLink pagination, server-side vs client-side paging~50 lines
references/docs/batching.md$batch endpoint, combining multiple requests, dependsOn sequencing~280 lines
references/docs/throttling.md429 handling, Retry-After, backoff strategy~90 lines
references/docs/errors.mdHTTP status codes, error response format, error codes~105 lines
references/docs/best-practices.md$select for performance, pagination, delta queries, batching~155 lines

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Research

learn-anything-in-one-hour

Teach users any new skill/knowledge X in ~1 hour using a fixed 4-step workflow optimized for complete beginners, focusing on 80/20 rule for maximum value in minimum time. Triggers when user asks to learn something new quickly, or mentions "learn X in one hour".

Archived SourceRecently Updated
Research

X/Twitter Research

# X/Twitter Research Skill

Archived SourceRecently Updated
Research

council

Convene the Council of High Intelligence — multi-persona deliberation with historical thinkers for deeper analysis of complex problems.

Archived SourceRecently Updated
Research

polymarket-openclaw-trader

Reusable Polymarket + OpenClaw trading operations skill for any workspace. Use when the user needs to set up, run, tune, monitor, and deploy an automated Polymarket trading project (paper/live), including env configuration, risk controls, reporting, and dashboard operations.

Archived SourceRecently Updated