home_mcp_gateway
A local MCP gateway that provides image generation and editing tools via external APIs (e.g., DALL-E) and local ComfyUI workflows, with authentication and policy enforcement.
README
home_mcp_gateway
home_mcp_gateway is a local MCP gateway for ZeroClaw and other MCP clients.
It exposes one local HTTP/SSE MCP endpoint, routes tool calls through a policy
engine, stores generated files as artifacts, and records jobs and audit events
in SQLite.
The important configuration rule is:
Users edit config/config.yaml.
config/config.example.yaml is the template/base configuration. For local use,
copy it to config/config.yaml. Both local Python runs and Docker Compose use
that same file. Root-level .env keeps tokens and provider secrets.
Quick Start
Requirements:
- Python 3.11 or newer.
PyYAML.
Create your local user config:
Copy-Item config/config.example.yaml config/config.yaml
Copy-Item .env.example .env
Edit config/config.yaml and .env, then run:
python -m pip install -e .
python -m app.main
The default local endpoint is http://127.0.0.1:8787.
Check the process:
Invoke-RestMethod http://127.0.0.1:8787/healthz
Invoke-RestMethod http://127.0.0.1:8787/readyz
Where User Settings Live
Use these files for different purposes:
| File | Purpose |
|---|---|
config/config.example.yaml |
Template/base config. Copy it to config/config.yaml. |
config/config.yaml |
Your local runtime config. Used by both Python and Docker Compose. It is git-ignored. |
.env.example |
Environment template. Copy it to .env for tokens and provider secrets. |
.env |
Local environment variables for Python and Docker Compose. It is git-ignored. |
tests/config/test.config.yaml |
Test-only config used by tests/run_tests.ps1. |
The application loads configuration in this order:
- Load root
.envinto the process environment without overriding existing variables. - Load
config/config.example.yaml. - If
CONFIG_PATHis set, deep-merge that YAML over the base config. - If
CONFIG_PATHis not set andconfig/config.yamlexists, deep-merge that file instead. - Replace placeholders such as
${IMAGE_API_KEY}with environment variables.
Common user-editable sections are:
server.host,server.portartifacts.root,artifacts.public_base_urldatabase.pathcallers.*.token_envpolicy.high_risk_allowed_callersmodules.image,modules.localimage
At the moment, only the image and localimage modules are actually used.
Other module directories may exist in the repository, but they are not part of
the current practical setup.
Secrets should stay in environment variables, not YAML files. Copy
.env.example to .env and put local token/provider values there.
For a custom config file outside the default location, set CONFIG_PATH
explicitly:
$env:CONFIG_PATH = "path/to/your.config.yaml"
python -m app.main
MCP Client Setup
Use the SSE transport URL:
[[mcp.servers]]
name = "home"
transport = "sse"
url = "http://127.0.0.1:8787/mcp"
deferred_loading = true
When the gateway is used from another Compose service on the same network, use:
[[mcp.servers]]
name = "home"
transport = "sse"
url = "http://home-mcp:8787/mcp"
deferred_loading = true
Available Tools
Built-in tools:
health_check: returns server and enabled-module status.artifact_get: returns artifact metadata and a signed download URL.artifact_get_image: compatibility helper that requires an inline-readable image artifact.job_status: returns a job visible to the caller.
Currently used module tools:
image_generate,image_edit: generate or edit images and store results as image artifacts.localimage_generate: generate images through a local ComfyUI workflow and store results as image artifacts.
For now, the practical module scope is limited to image and localimage.
Authentication And Policy
Callers authenticate with a bearer token:
Authorization: Bearer <token>
Tokens are compared against the environment variable named by each caller entry
in configuration. Anonymous callers can only use tools listed in
policy.anonymous_allowed_tools; by default that is health_check.
Artifacts and jobs are scoped to the caller that created them. Admin callers can
read shared artifacts only when their configured caller has
shared_artifact_read: true.
Docker
Docker Compose uses the same config/config.yaml as local Python runs:
Copy-Item config/config.example.yaml config/config.yaml
Copy-Item .env.example .env
# Edit config/config.yaml and .env.
docker compose up --build
Compose automatically reads .env, mounts config/config.yaml, and stores
artifacts plus SQLite metadata under ./artifacts.
When ZeroClaw also runs in Docker, artifact download_url values must point to
an address reachable from the ZeroClaw container. The gateway derives download
URLs from the address used to call MCP when possible, so a ZeroClaw MCP URL of
http://192.168.1.23:8787/mcp yields artifact URLs under
http://192.168.1.23:8787/artifacts. ARTIFACT_PUBLIC_BASE_URL remains the
fallback and defaults to http://127.0.0.1:8787/artifacts for host-local use.
Artifact download_url values are short-lived signed URLs, so clients can fetch
them directly without knowing the MCP Bearer token.
Tests
Tests use tests/config/test.config.yaml; that file is not the user config.
.\tests\run_tests.ps1
The test script sets CONFIG_PATH=tests/config/test.config.yaml and local test
tokens, then runs the Python unittest suite.
Documentation
- Developer documentation:
documents/ - Deployment notes:
deploy/README.md - Module extension notes:
dev_documents/module-extension.md - Chinese README:
README.zh-cn.md
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.