OWUI-Email-MCP
Enables drafting emails via Model Context Protocol in OpenWebUI using templates, generating short links for editable drafts without sending any emails.
README
OWUI-Email-MCP
Email drafts via MCP for OpenWebUI. Template-driven, plain text or HTML, nothing sent.
Email drafting for OpenWebUI over the Model Context Protocol. The model fills
a predefined template; the server stores the finished draft behind a short
link. With HTML_FORMAT=true the link downloads a ready-to-open .eml
file (HTML), otherwise it redirects straight to a mailto: draft (plain
text) — either way an editable draft in the user's own mail client. Nothing is
sent automatically, and the model never has to stream the whole email token
by token.
✨ Highlights
- Fast by design — the model outputs a ~10-token short link; the draft is built server-side and only lives behind that link
- Two formats, one switch —
HTML_FORMAT=true:.emldownload markedX-Unsent: 1that opens as an editable HTML draft (images, inline styles);HTML_FORMAT=false: instantmailto:redirect (plain text) - One click to the draft — either way the draft opens in the user's default mail client; links expire after a TTL
- Nothing is sent — the user's own mail client opens the draft; the user stays in control
- Multi-user by design — JWT auth against OpenWebUI's secret, per-user rate limiting
- Stateless-ish — drafts live in a TTL cache in memory (powered by cachetools); no database, no files
🚀 Setup
uv sync
cp .env.example .env
.env.example documents every setting; the ones you must set:
JWT_SECRET→ OpenWebUI'sWEBUI_SECRET_KEYPUBLIC_BASE_URL→ URL of this server as reachable from the user's browser (it is embedded in the short links), e.g.http://192.168.1.10:8000HTML_FORMAT→truefor HTML drafts via.emldownload,falsefor plain text via instantmailto:redirect
🏃 Run
uv run python main.py
The server listens on 0.0.0.0:8000:
| Endpoint | Auth | Purpose |
|---|---|---|
/mcp |
OpenWebUI JWT | MCP (streamable HTTP) for OpenWebUI's External Tools |
/m/{id} |
none | the draft: mailto: redirect (txt) or .eml download (html) |
/static/* |
none | static assets (e.g. the logo referenced by the email templates) |
In OpenWebUI: Admin Settings → External Tools → add server of type
MCP (Streamable HTTP) with URL http://<host>:8000/mcp and auth
Session. The model gets list_email_templates and compose_email and
is instructed to reply with the short link only.
🐳 Docker (optional)
Prebuilt images are published to ghcr.io on pushes to main (latest)
and on version tags (X.Y.Z):
docker run -d -p 8000:8000 \
--restart unless-stopped \
--env-file .env \
-v ./templates:/app/templates \
--name owui-email-mcp \
ghcr.io/th3r3alduk3/owui-email-mcp:latest
Or build the image locally: docker build -t owui-email-mcp . — the volume
mount keeps the templates editable without rebuilding the image.
🛠️ Tools
| Tool | Description |
|---|---|
compose_email |
fill a template, store the draft, return the short link |
list_email_templates |
list templates (name → template text) |
📝 Templates
Templates live in templates/ (name = filename without suffix) and
are read fresh on every call — edit them without restarting. HTML_FORMAT
selects which files are used:
HTML_FORMAT=false—*.txtfiles: first line = subject, blank line, then the body; the short link redirects to amailto:draftHTML_FORMAT=true—*.htmlfiles: the<title>is the subject; the short link downloads an.emldraft. Use inline styles only (a<style>block's CSS braces would clash withstr.format); images referenced assrc="/static/..."are rewritten to absolutePUBLIC_BASE_URLlinks so mail clients can load them{placeholders}are allowed anywhere and are filled viastr.formatwith the values from the compose call; literal braces are written{{and}}
⚠️ Limits
- Drafts are held in memory only and expire after
LINK_TTL(default 24 h); at mostMAX_STORED_EMAILSare kept. A restart clears all links. - The draft link is unauthenticated by design (the browser opening it has no OpenWebUI session): anyone with the unguessable link can read the draft until it expires. Don't put secrets in templates.
- MCP requests are rate-limited per user (
RATE_LIMIT_RPS/RATE_LIMIT_BURST), keyed on the JWT'sidclaim; tool errors never leak internals (mask_error_details). - The server speaks plain HTTP — put it behind a reverse proxy for TLS.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.