MailNet MCP Server
Provides tools to manage Gmail emails including send, read, search, label, archive, reply, delete, and draft.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@MailNet MCP ServerShow me my unread emails from today."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MailNet MCP Server
MailNet Server is a unified, agentic email orchestration server built for the Model Context Protocol. It supports Gmail and Outlook with standardized metadata, secure credential injection, and a rich toolset for assistant-driven workflows. It is the MCP server that powers MailNet Mailing Agentic AI, live at getmailnet.com.
Features
Unified Gmail + Outlook abstraction
Automatic token refresh and credential hygiene
Standardized base class for provider extension
Agentic email settings endpoint (tone, signature, thread context, etc.)
Modular toolset: send, read, search, label, archive, reply, delete, draft
Related MCP server: Mailing Manager MCP
How it fits into MailNet
In the deployed MailNet stack this server runs as its own container (mcp, streamable HTTP on port 9111) and is the only component that talks to mailboxes. Everything email-shaped goes through it:
LangGraph agent (api) ──┐
├──> MCP server :9111 ──> provider layer ──> Gmail API
Scheduler (APScheduler) ─┘ (this repo) └──> Microsoft GraphPer-request credentials, no stored state. The agent and the scheduler pass each user's OAuth tokens as encrypted HTTP headers on every call; the server decrypts, refreshes if needed, acts, and forgets. It holds no user database.
One tool surface, two providers. Tools like
send_email,read_emails, andsearch_emailsaccept an optional provider parameter; the provider layer hides the differences between the Gmail API and Microsoft Graph.Standalone by design. The same server also runs solo for local/Claude Desktop use with token files, as described below — MailNet is one consumer of it, not a requirement.
Installation
Running the full MailNet application (web UI, agent, and this server together)? Follow the MailNet repo's Getting started instead; it wires this server up via Docker Compose automatically. The steps below are for running the MCP server standalone (for example with Claude Desktop).
1. Manual Clone & Launch
git clone https://github.com/Astroa7m/MailNet-MCP-Server.git
cd MailNet-MCP-ServerInstall requirements
pip install -r requirements.txt
Note if you are going to use uv for launching you should first install it via:
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
Launch
Either with uv via:
uv run -m mcp_launcher.server
or
python -m mcp_launcher.server
Acquiring Azure Token (Personal local use)
Go to
email_client/outlook_helpers.py.run the file (by default runs
acquiring_azure_token_for_personal_usefunction. To acquireclient_id&client_secretPlease check Azure Authorization Guide And make sure to add them to your env vars with the names shown below.
def acquiring_azure_token_for_personal_use():
load_dotenv()
client_id = os.getenv("AZURE_APPLICATION_CLIENT_ID")
client_secret = os.getenv("AZURE_SECRET_VALUE")
OutlookClient(client_id=client_id, client_secret=client_secret,
redirect_uri="http://localhost:3000/callback")It will do the following:
launch the browser and prompt you to sign in to your outlook account.
After successful login and approval of permissions, it will redirect you to
http://localhost:3000/callbackunless you specified different url in the constructor viaredirect_uriparam.Copy the code after
code=and before&client_infowithin the browser url and paste it in the terminal where you launched the file.Done now you will have your azure token under
email_clientnamedazure_token.jsonby default (can be changed viatoken_file_nameparam inOutlookClientconstructor).
Provide that path to
AZURE_PREFERRED_TOKEN_FILE_PATHenv variable and you are good to go.
Acquiring Google Token (Personal local use)
Go to
email_client/gmail_helpers.py.run the file (by default runs
acquiring_google_token_for_personal_usefunction. To acquiregoogle_credentials_filePlease check Google Authorization Guide And make sure to add the path to it in your env vars with the name shown below.
def acquiring_google_token_for_personal_use():
load_dotenv()
google_credentials_file = os.getenv("GOOGLE_CREDENTIALS_FILE_PATH")
GmailClient(credential_file=google_credentials_file)It will do the following:
launch the browser and prompt you to sign in to your gmail account.
After successful login and approval of permissions, it will redirect you to a window with a message
The authentication flow has completed. You may close this window.that means you are done here.Done now you will have your google token under
email_clientnamedgoogle_token.jsonby default (can be changed viatoken_file_nameparam inGmailClientconstructor).
Provide that path to
GOOGLE_CREDENTIALS_FILE_PATHenv variable and you are good to go.
Environment Variables
Check the Azure Authorization Guide and Google Authorization Guide to learn how to set up both accounts and get your credentials ready.
Set your environment variables for provider credentials:
Gmail
GOOGLE_CREDENTIALS_FILE_PATH=path/to/google_credentials.json
GOOGLE_PREFERRED_TOKEN_FILE_PATH=path/to/google_token.json
Outlook (Azure)
AZURE_APPLICATION_CLIENT_ID=your-client-id
AZURE_CLIENT_SECRET_VALUE=your-secret
AZURE_PREFERRED_TOKEN_FILE_PATH=path/to/azure_token.json
Other (Important for local use)
The following env variable is important to be set when running it locally or for Claude Desktop. The value deosn't matter but as long as the field is there you will be able to run it locally. It was introduced to make the server flexible to be run over http/s or stdio and to route the server to either look for credentials in local files or expect it from the client via http/s headers.
is_local="true"
Claude Desktop Integration
Add the following to your claude_desktop_config.json:
{
"mcpservers": {
"email_mcp": {
"command": "uv",
"args": [
"--directory",
"C:\\Path\\To\\mcp-server",
"run",
"-m",
"mcp_launcher.server"
],
"env": {
"AZURE_APPLICATION_CLIENT_ID": "<AZURE_APPLICATION_CLIENT_ID>",
"AZURE_CLIENT_SECRET_VALUE": "<AZURE_CLIENT_SECRET_VALUE>",
"AZURE_PREFERRED_TOKEN_FILE_PATH": "C:\\Path\\To\\azure_token.json",
"GOOGLE_CREDENTIALS_FILE_PATH": "C:\\Path\\To\\google_credentials.json",
"GOOGLE_PREFERRED_TOKEN_FILE_PATH": "C:\\Path\\To\\google_token.json",
"is_local": "true"
}
}
}
}Agentic Email Settings
You can view settings via the load_email_settings tool or update them via the update_email_settings tool.
{
"language": "en",
"tone": "formal",
"writing_style": "clear_and_concise",
"sender_name": "Ahmed Samir",
"organization_name": "Kalima Tech",
"include_signature": true,
"signature": "Best regards,\n{{sender_name}}\n{{organization_name}}",
"preferred_greeting": "Dear {{recipient_name}},",
"auto_adjust_tone": true,
"include_thread_context": true,
"character_limit": 1000,
"prompt_prefix": "You are an AI email assistant for {{organization_name}}. Keep messages professional, polite, and to the point.",
"default_provider": "google"
}Tools Supported
Tool | Description |
send_email | Compose and send messages |
read_email | Fetch and inspect messages |
create_draft | Prepare messages |
send_draft | Finalize and send |
search_email | Query inbox with semantic filters |
toggle_label | Modify categories/labels |
archive_email | Clean up inbox |
reply_email | Respond in thread context |
delete_email | Remove messages |
load_email_settings | View current email settings |
update_email_settings | Update runtime email settings |
Contributing
MailNet server is modular and extensible. To add a new provider, subclass the base client and implement the predefined hooks. PRs welcome!
This server cannot be deployed
Maintenance
Related MCP Connectors
Stateful email for AI agents — read inboxes, reply in-thread, draft with approval.
Email for AI agents — send, receive as a webhook, manage domains, templates, routing.
Email infrastructure for AI agents — send, receive, search, and reply to email over MCP.
Real email inboxes for AI agents: create inboxes, catch verification codes, extract OTPs, reply.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage Gmail and Outlook inboxes through natural language, supporting email search, reading, and reply operations.-
- AlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage multiple email accounts with secure credentials, local full-text search, thread-aware replies, and automation.6 npmMIT
- AlicenseNot gradedqualityCmaintenanceEnables unified email management across Gmail, Outlook, iCloud, and IMAP providers with tools for search, send, organize, and batch operations via natural language.779 npmMIT
- AlicenseBqualityBmaintenanceUniversal email integration for AI assistants, supporting any IMAP provider with 18 tools for reading, sending, managing, and automating emails.2813 npmMIT