Homebox MCP Server
An MCP server that exposes Homebox inventory management as tools for AI agents, enabling search, add, update, delete products, locations, tags, and maintenance entries.
README
Homebox MCP Server
An MCP server that exposes Homebox inventory management as tools for AI agents.
What it does
This server lets an AI agent (Claude, Hermes, OpenClaw, or any MCP-compatible client) read and write your Homebox inventory: search, add, update, and delete products and locations, manage tags and maintenance entries — all through the unified Homebox v0.26+ entities API.
It runs in two transports:
- stdio — for local agents that launch the server as a subprocess.
- Streamable HTTP (
/mcp) — for remote agents that talk HTTP. Per-request credentials are passed via request headers.
Key behaviors:
- Targets Homebox v0.26.0+ (the entity-merge release) — items and
locations share
/api/v1/entities, withparentIddenoting the parent entity. Older Homebox releases using/api/v1/itemsand/api/v1/locationsare not supported. search_productswith onlynamesplits the phrase into terms and queries each in parallel for broader recall.add_productandupdate_product_by_iduse grouped parameter objects (identity,location,identifiers,purchase,metadata,status) so large calls stay readable.- Maintenance tools (
create_maintenance_entry_for_product,list_maintenance_entries,update_maintenance_entry,delete_maintenance_entry) operate against/api/v1/maintenance. - Multi-collection (tenant) support is exposed via the Homebox-native
X-Tenantrequest header on the HTTP transport.
Installation
Requirements
| Component | Version |
|---|---|
| Homebox (server) | v0.26.0 or newer (tested v0.26.2) |
| Python | 3.10.13+ |
| Dependencies | Listed in pyproject.toml |
uv (recommended) |
https://docs.astral.sh/uv/ |
Older Homebox releases that still expose /api/v1/items and
/api/v1/locations are not supported — the server talks to
/api/v1/entities exclusively.
Install the server
git clone <repo-url> homebox-mcp
cd homebox-mcp
uv sync # installs runtime + dev dependencies
cp .env.example .env # then edit .env with your Homebox credentials
.env.example:
HOMEBOX_BASE_URL=https://your-homebox-instance.com
HOMEBOX_USERNAME=your_email@example.com
HOMEBOX_PASSWORD=your_password
# Optional HTTP server settings (server_http.py)
# FASTMCP_HOST=0.0.0.0
# FASTMCP_PORT=8000
Docker
docker build -t homebox-mcp .
docker run -p 8000:8000 \
-e HOMEBOX_BASE_URL="https://your-homebox-instance.com" \
-e HOMEBOX_USERNAME="your_email@example.com" \
-e HOMEBOX_PASSWORD="your_password" \
homebox-mcp
Run the server
just start # stdio mode (default for local agents)
just start-http # Streamable HTTP mode at http://0.0.0.0:8000/mcp
Or directly:
uv run main.py # stdio
uv run server_http.py # streamable HTTP
Install inside an agent
Pick the snippet that matches your agent and copy-paste it into the agent's
MCP configuration. Replace the values in <…> with your own.
The examples assume stdio mode (local agent) with credentials read from
environment / .env. Streamable HTTP mode exposes the same tools but takes
credentials through request headers instead.
Claude Desktop / Claude Code (claude_desktop_config.json)
{
"mcpServers": {
"homebox": {
"command": "uv",
"args": ["--directory", "/absolute/path/to/homebox-mcp", "run", "main.py"],
"env": {
"HOMEBOX_BASE_URL": "https://your-homebox-instance.com",
"HOMEBOX_USERNAME": "your_email@example.com",
"HOMEBOX_PASSWORD": "your_password"
}
}
}
}
Hermes (stdio)
Add to your Hermes MCP servers configuration:
{
"homebox": {
"command": "uv",
"args": ["--directory", "/absolute/path/to/homebox-mcp", "run", "main.py"],
"env": {
"HOMEBOX_BASE_URL": "https://your-homebox-instance.com",
"HOMEBOX_USERNAME": "your_email@example.com",
"HOMEBOX_PASSWORD": "your_password"
}
}
}
OpenClaw (stdio)
{
"servers": {
"homebox": {
"command": "uv",
"args": ["--directory", "/absolute/path/to/homebox-mcp", "run", "main.py"],
"env": {
"HOMEBOX_BASE_URL": "https://your-homebox-instance.com",
"HOMEBOX_USERNAME": "your_email@example.com",
"HOMEBOX_PASSWORD": "your_password"
}
}
}
}
Remote / Streamable HTTP
If you run the server with uv run server_http.py (or via the Docker image),
any MCP-compatible HTTP client can reach it at http://host:8000/mcp and pass
credentials via request headers:
HOMEBOX_USERNAME: your_email@example.com
HOMEBOX_PASSWORD: your_password
X-Tenant: 00000000-0000-0000-0000-000000000000 # optional, multi-collection
HOMEBOX_BASE_URL is read only from the runtime environment of the server
process; it cannot be overridden per request.
Finding the X-Tenant value
Open Homebox, log in, then in DevTools → Network find the request
GET /api/v1/users/self and copy the X-Tenant request header. Pass it on
every MCP request via the X-Tenant header.
Available tools
17 tools are exposed. Grouped parameter objects accept both snake_case and camelCase keys.
Products (entities)
search_products
Search products via Homebox query plus local filters. Provide at least one
filter; name is split into per-term queries when no other strict filter is
provided.
name(Optional[str]) — product name (or phrase) queryitem_id(Optional[str]) — exact product UUIDasset_id(Optional[str]) — asset ID filterserial_number(Optional[str])model_number(Optional[str])manufacturer(Optional[str])query(Optional[str]) — generic full-text query
Returns: list of matching product dicts, or an error string.
add_product
Create a new product or subitem with grouped parameters. If both location
and parent_item are provided, parent_item wins (subitem creation).
identity(dict, required):{"name", "description", "asset_id"}location(Optional[dict]):{"id"}or{"name"}— for standalone productsparent_item(Optional[dict]):{"id"}or{"name"}— for subitemsidentifiers(Optional[dict]):{"serial_number", "model_number", "manufacturer"}purchase(Optional[dict]):{"price", "seller", "date", "warranty_expires"}metadata(Optional[dict]):{"notes", "quantity", "tag_ids", "tag_names", "custom_fields"}status(Optional[dict]):{"archived", "insured", "lifetime_warranty"}
Returns: created product dict (after optional update pass), or error string.
update_product_by_id
Update an existing product. Pass IDs (item_id, location.id) — names are
only resolved through list_locations outside this tool.
item_id(str, required) — UUID of the item to updateidentity(Optional[dict]):{"new_name", "description", "asset_id"}location(Optional[dict]):{"id"}or{"name"}identifiers(Optional[dict]):{"serial_number", "model_number", "manufacturer"}purchase(Optional[dict]):{"price", "seller", "date", "warranty_expires"}metadata(Optional[dict]):{"notes", "quantity", "tag_ids", "tag_names", "custom_fields"}status(Optional[dict]):{"archived", "insured", "lifetime_warranty"}
Returns: updated product dict, or error string.
add_bulk_products
Add multiple products in one call. Each entry uses the grouped shape from
add_product.
items(list[dict], required) — each item must contain anidentity.name- Returns: list (one result per input item).
delete_product_by_id
Delete a single product by ID.
item_id(str, required) — UUID of the itemproduct_name(str, required) — used for the success/error message only
Returns: success message, or error string.
delete_bulk_products
Delete multiple products. Delegates to delete_product_by_id for each entry.
items(list[dict], required) — each entry{item_id, product_name}- Returns: list (one result message per input item).
Locations
list_locations
Return Homebox's location tree (GET /api/v1/entities/tree).
- No parameters.
- Returns: list of top-level tree nodes (each with nested
children).
create_location
Create a new location (entity with isLocation implied by being parent-less).
name(str, required)description(Optional[str])parent_id(Optional[str]) — UUID of a parent location
Returns: created location dict, or error string.
update_location_by_id
Update a location's editable fields. Fetches the current state first to preserve omitted fields.
location_id(str, required) — UUID of the locationnew_name(Optional[str])new_description(Optional[str])new_parent_id(Optional[str]) — pass empty string""to detach from parent
Returns: updated location dict, or error string.
delete_location_by_id
Delete a single location.
location_id(str, required) — UUID of the location- Returns: success message, or error string.
list_items_by_location
List direct child items of a location (uses ?parentIds=…).
location_id(str, required) — UUID of the location- Returns: list of
{"id", "name"}items, or error string.
Tags
list_tags
Return all tags from /api/v1/tags.
- No parameters.
- Returns: list of
{"id", "name"}tags, or error string.
create_tag
Create a tag if it does not already exist (case-folded de-dupe).
name(str, required)- Returns: created or pre-existing tag dict, or error string.
Maintenance
create_maintenance_entry_for_product
Create a maintenance entry for an existing product. Provide
scheduled_date, completed_date, or both; if neither, both default to
today.
product_id(Optional[str]) — preferred if knownproduct_name(Optional[str]) — resolved to a product ID ifproduct_idis missingname(str, required) — title/summary of the taskdescription(Optional[str])cost(Optional[float])scheduled_date(Optional[str]) —YYYY-MM-DDcompleted_date(Optional[str]) —YYYY-MM-DD
Returns: created maintenance entry dict, or error string.
list_maintenance_entries
List maintenance entries from Homebox. The global /api/v1/maintenance
endpoint is broken on Homebox v0.26+; this tool falls back to enumerating
per-entity maintenance lists and aggregating them.
product_id(Optional[str]) — restrict to entries for one productproduct_name(Optional[str]) — used to resolve the entity whenproduct_idis missing- Returns: list of maintenance entry dicts, or error string.
update_maintenance_entry
Update an existing maintenance entry. Homebox v0.26+ exposes maintenance only per-entity, so the owning product must be supplied.
entry_id(str, required) — UUID of the entryproduct_id(Optional[str]) — UUID of the owning entity (preferred)product_name(Optional[str]) — resolved to an entity whenproduct_idis missingname(Optional[str])description(Optional[str])cost(Optional[float])scheduled_date(Optional[str]) —YYYY-MM-DDcompleted_date(Optional[str]) —YYYY-MM-DD
Returns: updated maintenance entry dict, or error string.
delete_maintenance_entry
Delete a maintenance entry.
entry_id(str, required) — UUID of the entry- Returns: success message, or error string.
Testing
The project ships unit tests and end-to-end tests.
Unit tests (no Docker required)
just test
# or: uv run pytest tests/ -v -k "not e2e"
End-to-end tests (requires Docker)
The E2E suite spins up a pinned Homebox v0.26.2 container, builds the MCP server image, registers a test user, and exercises all 17 tools through the Streamable HTTP transport. Each tool's effect is then verified against the real Homebox API (not just the MCP response envelope).
just test-e2e
What it does:
docker compose -f docker-compose.e2e.yml up -d— start Homebox v0.26.2 on port31745and the MCP server on port31746.- Wait for the Homebox healthcheck.
- Register a test user (idempotent).
- Run
pytest tests/e2e/ -m e2e. - Tear down all containers.
See docker-compose.e2e.yml for the pinned image and tests/e2e/ for the
suite. The suite covers every tool:
| Category | Tools tested |
|---|---|
| Locations | list_locations, create_location, update_location_by_id, list_items_by_location, delete_location_by_id |
| Products | search_products, add_product, update_product_by_id, add_bulk_products, delete_product_by_id, delete_bulk_products |
| Tags | list_tags, create_tag |
| Maintenance | create_maintenance_entry_for_product, list_maintenance_entries, update_maintenance_entry, delete_maintenance_entry |
Recommended workflow for adding items
- Find a location for a standalone product — call
list_locationsto get a location UUID. - Find a parent for a subitem — call
search_products(name=...)to resolve the parent item UUID. - Create or update — use
add_product(new) orupdate_product_by_id(existing). If bothlocationandparent_itemare passed toadd_product,parent_itemis used andlocationis ignored for the creation target.
Justfile commands
just --list # Show all commands
just start # Start MCP server in stdio mode
just start-http # Start MCP server in Streamable HTTP mode
just stop # Stop running HTTP servers
just lint-format # Run linting checks
just lint-fix # Run linting and auto-fix issues
just test # Run unit tests
just test-e2e # Run E2E tests (requires Docker)
just wipe # Clean cache/build artifacts
just docker-build # Build Docker image
just docker-run # Run Docker container
Repository layout
homebox-mcp/
├── main.py # FastMCP server, all 17 tools, helpers
├── models.py # Grouped dataclasses for tool inputs
├── server_http.py # Streamable HTTP transport entrypoint
├── request_config_middleware.py # Per-request credential/tenant scoping
├── Dockerfile # HTTP server image
├── docker-compose.e2e.yml # Pinned Homebox + MCP server stack
├── justfile # Dev/test commands
├── scripts/e2e-setup.sh # Test user registration helper
├── tests/
│ ├── test_main.py # Unit tests (mocked HTTP)
│ ├── test_request_config_middleware.py
│ ├── test_transport_servers.py
│ └── e2e/ # E2E tests against real Homebox v0.26.2
└── pyproject.toml
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.