Skip to main content
Glama
uaixo

MCP Server for Multiple Outlook Accounts

by uaixo

MCP Server for Multiple Outlook Accounts

A local, single-connector Model Context Protocol server that lets an AI assistant operate several Outlook / Microsoft 365 mailboxes at once โ€” searching, reading, drafting, sending, and organising mail โ€” through one safety-annotated tool surface, with OAuth tokens that never leave your machine.

This is the Outlook / Microsoft Graph provider variant of a provider-neutral specification.


Project status: feature-complete (offline build)

Built with TypeScript 6.0.3; build, typecheck, tests (179), and format all green. All eight capabilities are implemented and the offline build is feature-complete. What exists today:

  • โœ… Architecture design + requirements traceability matrix (doc/).

  • โœ… Auth core: Entra credential-source discovery, MSAL public-client (consent + silent refresh), secure token store (0600/0700, atomic + cross-process-locked), account-selection registry.

  • โœ… Account-management CLI: outlook-mcp-auth connect | list | remove.

  • โœ… Microsoft Graph client: thin fetch wrapper with a per-request timeout, bounded jittered retry (with the no-duplicate-send policy), and actionable error mapping.

  • โœ… Read tools: list_accounts (C1), search_conversations (C2), read_conversation (C3) โ€” with Gmail-style search-operator translation and bounded, truncating output.

  • โœ… Write tools: create_draft (C4) and send_message (C5) โ€” recipient parsing (Display Name <addr>), header-injection stripping, allow-listed/TOCTOU-safe attachments, local outgoing-size validation, and reply threading. Small attachments ride inline; large ones (> ~3 MB) upload to the draft via a Graph upload session, so send becomes create-draft โ†’ upload โ†’ send with the final /send under the nonDuplicable retry policy โ€” a retry can never double-deliver.

  • โœ… Organise tools: list_labels (C6), create_label (C7), and organize_mail (C8) โ€” the label-decomposition fan-out that maps one neutral organise request to the right mix of Graph category-PATCH / read-state / move calls (archive, trash, junk โ€” mutually exclusive), applied per message across a conversation under a bounded concurrency limit, reporting the union of resulting labels. list_labels enumerates folders recursively (full paths).

  • โœ… Hardening + onboarding: a secret-redaction boundary so tokens/credentials can never reach the logs (NFR-SEC-6), and the operator onboarding guide (Entra app registration, unverified-app consent, the connect CLI).

Remaining: the live acceptance runs (real browser consent + real Graph calls) that the operator performs against an Entra app registration + Outlook mailbox. See doc/architecture.md ยง13.

Tests mock Microsoft Graph and MSAL. Live ยง13 acceptance โ€” real browser consent and Graph calls โ€” requires an Entra app registration + Outlook mailboxes and is run locally by the operator.


Related MCP server: msgraph-mcp

Capabilities (spec ยง5)

Tool

Purpose

Destructive?

Status

list_accounts

List connected mailboxes

No

โœ… live

search_conversations

Search a mailbox (paged)

No

โœ… live

read_conversation

Read a full conversation

No

โœ… live

create_draft

Compose a draft (not sent)

No

โœ… live

send_message

Send immediately

Yes

โœ… live

list_labels

List categories + folders

No

โœ… live

create_label

Create a category/folder

No

โœ… live

organize_mail

Tag / move / read-state

Yes

โœ… live

Plus an out-of-band account-management CLI (outlook-mcp-auth connect | list | remove).


Development

Requires Node.js โ‰ฅ 18 (developed on Node 22).

npm install
npm run typecheck      # tsc --noEmit
npm run build          # compile to dist/
npm test               # vitest (Graph/MSAL mocked)
npm run format:check   # prettier

Configuration

All operational knobs are environment variables (see .env.example):

Env var

Meaning

Default

OUTLOOK_MCP_DATA_DIR

tokens + app-registration configs

~/.outlook-mcp

OUTLOOK_OAUTH_CREDENTIALS

pin one app registration (disables discovery)

unset

OUTLOOK_MCP_ATTACHMENTS_DIR

allow-list for path attachments

unset (disabled)

OUTLOOK_MCP_LOCK_TIMEOUT_MS

token-store lock wait

12000

OUTLOOK_MCP_REQUEST_TIMEOUT_MS

per-Graph-call timeout

30000

Connecting a mailbox

For the full walkthrough โ€” Entra app registration, the unverified-app consent policy, and troubleshooting โ€” see doc/ONBOARDING.md. The short version:

  1. Register a public-client app in Entra ID with the redirect URI http://localhost and the delegated scopes Mail.ReadWrite, Mail.Send, User.Read, offline_access.

  2. Drop a credentials*.json into the data dir (or point OUTLOOK_OAUTH_CREDENTIALS at it):

    { "clientId": "<application-client-id>", "tenant": "common" }

    Multiple credentials*.json files are auto-discovered, so accounts under different app registrations each refresh with the client that authorised them.

  3. Connect, list, and remove mailboxes:

    outlook-mcp-auth connect            # opens the browser for consent
    outlook-mcp-auth connect --source acme   # pick a specific app registration
    outlook-mcp-auth list
    outlook-mcp-auth remove user@example.com

Security posture

OAuth tokens are stored only on the local machine (file mode 600 in a 700 data dir). Reading local files by path for attachments is disabled by default and only allowed from an explicit allow-list. The server never logs tokens, credentials, or message content. See doc/architecture.md ยง9.

License

MIT โ€” see LICENSE.

Available Tools

1 tool
list_accountsList connected accountsA
Read-onlyIdempotent

List the Outlook / Microsoft 365 mailboxes connected to this server. Use the returned identities as the 'account' selector for other tools.

ParametersJSON Schema
NameRequiredDescriptionDefault

No parameters

TDQS

A4.7/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already indicate readOnly and idempotent; description adds context about output usage but not additional behavioral traits.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Two sentences efficiently convey purpose and usage without any extraneous information.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given zero parameters and rich annotations, the description fully explains what the tool does and how to use its output.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

No parameters exist, and schema coverage is 100%, so baseline score of 4 applies; description adds no additional parameter info but none is needed.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states it lists Outlook/Microsoft 365 mailboxes and explains the purpose of the output.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly tells the agent to use the returned identities as the 'account' selector for other tools, guiding correct usage.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 1 tool updatev0.1.0
    • First observedlist_accounts

TDQS

A4.2/5.0

Scored across 1 tool

Disambiguation5/5

With only one tool, there is no possibility of ambiguity. Agents will always select list_accounts when any tool is needed.

Naming Consistency5/5

The single tool follows a clear verb_noun pattern (list_accounts), making its purpose immediately understandable.

Tool Count1/5

A single listing tool is far too few for a server purporting to handle multiple Outlook accounts. The user would expect tools for email operations, calendar management, or at least sending/reading messages.

Completeness1/5

The server only lists accounts and provides no way to actually interact with them. There are obvious missing operations like read_email, send_email, search_mailbox, or manage folders, leaving the surface severely incomplete.

Maintenance

ActivityStale
ResponsivenessNo issues

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Connectors

Related MCP Servers

  • F
    license
    A
    quality
    C
    maintenance
    Local-first MCP server for agents that need to work across multiple Gmail and Microsoft 365 accounts without cloud token storage.
    6
    -
  • F
    license
    B
    quality
    B
    maintenance
    MCP server providing AI assistants with full access to Microsoft Outlook email and calendar via the Microsoft Graph API, featuring 26 tools for mail, calendar, contacts, and scheduling with delegated authentication.
    29
    -
  • A
    license
    Not graded
    quality
    D
    maintenance
    Lightweight MCP server that enables AI agents to manage Microsoft Outlook email, calendar, and files using PKCE authentication with only a Client ID.
    MIT
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables reading, sending, and managing Microsoft 365/Outlook emails through MCP tools with OAuth 2.1 authentication.
    92
    MIT