AItomation MCP relay
Enables external AI applications to access MCP tools by providing a user-hosted HTTPS endpoint that forwards requests to a paired desktop client.
README
AItomation MCP relay
User-hosted HTTPS edge for inbound AI connectors. Other apps call this process on an address you own. The app on your computer pairs to it later and runs the tools. This process never runs tools itself.
This repository is the connector process only. It is not the desktop app and it is not the Team Space team server.
We do not host this door for you. Place it on a server you control. A
desktop app update does not update this process. A desktop v* tag does
not publish this image.
One process, one project
One relay talks to one device and one project.
- Three projects that need a public HTTPS door need three processes.
- Path prefixes on one name (
/project-a/mcp) are refused. - The Team Space server on the same machine is a fourth process. Same machine is fine. Same process is not.
Quick start (Docker)
docker pull ghcr.io/stanislavmandrik621/aitomation-mcp-relay:latest
docker run --rm -p 8790:8790 \
-e MCP_RELAY_HOST=0.0.0.0 \
-e MCP_RELAY_PORT=8790 \
-e MCP_RELAY_DATA_DIR=/data \
-v mcp-relay-data:/data \
ghcr.io/stanislavmandrik621/aitomation-mcp-relay:latest
Default listen port is 8790. Do not use 8787, 8788, or 8789.
GET / and GET /health answer 200. A browser shows a short status
page. curl and a later pair probe (Accept: text/plain) still get the
one-line text AItomation MCP relay.
POST /mcp without a sign-in token returns 401 plus a
WWW-Authenticate header that points at the well-known JSON. After
you pair the app and complete OAuth, that POST is forwarded to the
app. Tools never run on this process. Pairing uses /pair/hello,
/pair/next, /pair/reply, and /pair/unavailable.
Official images are multi-arch (linux/amd64 and linux/arm64). Prefer
pinning a version tag in production.
Public certificate on a public host
Cloud hosts that call this connector (ChatGPT, Claude, Grok, Gemini
Enterprise) need a certificate from a public certificate authority.
Do not use a self-made certificate on a public host. Put a reverse
proxy with a public certificate in front of this process, or pass
MCP_RELAY_TLS_CERT_FILE and MCP_RELAY_TLS_KEY_FILE that already
come from that authority.
Environment
| Name | Default | Notes |
|---|---|---|
MCP_RELAY_HOST |
127.0.0.1 |
Docker sets 0.0.0.0. |
MCP_RELAY_PORT |
8790 |
Reserved ports are refused. |
MCP_RELAY_DATA_DIR |
./data |
Use an absolute path under a service manager. |
MCP_RELAY_PAIR_SECRET |
unset | Pair secret stays in env, never in argv. |
MCP_RELAY_PUBLIC_URL |
unset | Public https origin (optional /mcp). Used in well-known resource. |
MCP_RELAY_STATIC_HEADER_HASH |
unset | Optional SHA-256 hex of the Claude request-header key for the paired project. |
MCP_RELAY_TLS_CERT_FILE |
unset | Optional. Public CA for a public host. |
MCP_RELAY_TLS_KEY_FILE |
unset | Pair with the cert file. |
This process does not read Team Space environment names.
OAuth stays on port 8790. Authorization URL is /authorize. Token
URL is /token. Include offline_access in scopes. Gemini Enterprise
uses the confidential client id minted on first start (secret is written
once under the data dir) and redirect
https://vertexaisearch.cloud.google.com/oauth-redirect. Well-known
JSON is at /.well-known/oauth-protected-resource and
/.well-known/oauth-protected-resource/mcp. There is no HTML catch-all
on those paths.
Alternate (clone + npm)
git clone https://github.com/stanislavmandrik621/aitomation-mcp-relay.git
cd aitomation-mcp-relay
npm ci
npm start
npm start is for development. For a real service, build once
(npm run build) and run node dist/server.js under a service manager.
Step-by-step service unit, three-project compose, and HTTPS proxy snippets: docs/SELF-HOST.md.
Stop
SIGTERM or SIGINT stops new work, waits for in-flight POST /mcp
forwards to finish, then exits. The process does not kill itself
mid-POST.
Source of truth
Day-to-day edits live in the AItomation monorepo packages/mcp-relay
and are published here for customers. Official GHCR write is this
repo's publish-image workflow. The desktop cut does not publish
this image.
License
MIT - see LICENSE.
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.