MeuralMCP
Provides MCP tools for coding agents to inspect and control Meural Canvas devices on the local network, including listing devices, getting status, and setting the current image.
README
MeuralMCP
Small LAN-first daemon, REST API, and MCP server for keeping Meural Canvas previews loaded.
Features
- Quiet daemon loop that retries often-failing frames without noisy logs.
- REST API protected by a shared token.
- Hosted remote MCP endpoint for coding agents, protected by the same shared token.
- Optional stdio MCP adapter for clients that cannot connect to remote MCP directly.
- Per-device current image storage.
- Orientation validation before writing a preview.
- In-memory/status tracking for LAN reachability failures.
- Optional cloud init that creates/assigns one-item blank galleries for landscape and portrait devices, sets hold durations, and syncs configured cloud device IDs.
Credits
This project builds on public Meural local-control work from the community:
- Guy Sie's HA-meural, introduced on the Home Assistant forum, demonstrated that Meural Canvas devices expose both a cloud API and a local interface suitable for Home Assistant control.
- NETGEAR's support article
documents the built-in browser-accessible
/remotecontroller on the same LAN. - bigboxer23/meural-control documented the practical issue with transient preview/postcard display and the more reliable single-playlist/gallery pattern.
Quick Start
Most users should run cloud init first. It is the step that discovers your frames, writes the starter config, changes Meural cloud timeout settings, and assigns the neutral blank gallery used by the local preview manager.
Cloud init changes these Meural cloud values:
- Sets
imageDuration=86400,previewDuration=86400, andoverlayDuration=120on discovered devices. - Finds or creates one landscape and/or one portrait blank gallery.
- Uploads one blank image to each blank gallery if the gallery is empty.
- Assigns each device to the matching blank gallery for its orientation.
- Syncs the devices after changing these settings.
Cloud init does not delete your existing Meural images or galleries, but it can
change which gallery is assigned to each discovered device. Review the generated
config.json after init and edit device names, IPs, orientations, or enabled
flags before starting the daemon.
python -m venv .venv
. .venv/bin/activate
pip install -e .
meural-mcp --storage-dir ~/.config/meural-mcp init-cloud \
--username "$MEURAL_USERNAME" \
--password "$MEURAL_PASSWORD"
Package dependencies are declared in pyproject.toml; pip install -e . installs
the CLI plus bounded runtime dependencies for FastAPI, Uvicorn, Requests, and MCP.
llms.txt is included as a concise map for coding agents and documentation
ingestion. AGENTS.md contains contributor guardrails for keeping this repo
public-safe.
If you do not want cloud init, init-local can create an empty generic config:
meural-mcp --storage-dir ~/.config/meural-mcp init-local --api-token "change-me"
With init-local, you must manually fill in device names, cloud IDs, LAN IPs,
orientations, and enabled flags in ~/.config/meural-mcp/config.json. It also
does not set timeout values or blank galleries in Meural cloud.
Run the API:
meural-mcp --storage-dir ~/.config/meural-mcp serve --host 127.0.0.1 --port 8733
Keep the REST API bound to localhost where possible. If you want to expose it to other machines, put nginx or Caddy in front of it and terminate HTTPS there rather than binding the API directly to the LAN as plain HTTP.
Check summary status:
curl http://127.0.0.1:8733/status \
-H "Authorization: Bearer change-me"
Run the daemon:
meural-mcp --storage-dir ~/.config/meural-mcp daemon
Services and HTTPS
Sample service and reverse-proxy files are provided in examples/:
examples/systemd/meural-mcp-daemon.serviceexamples/systemd/meural-mcp-api.serviceexamples/systemd/meural-mcp-user.serviceexamples/systemd/meural-mcp-api-user.serviceexamples/reverse-proxy/Caddyfileexamples/reverse-proxy/nginx.conf
init-cloud is a one-shot CLI setup command. Run it manually so you can see and
approve the cloud changes. The daemon and API are separate long-running
processes: the daemon keeps previews loaded, and the API serves local status and
image writes. In the system service examples, both long-running services use
/var/lib/meural-mcp for config/state and expect the project installed at
/opt/meural-mcp/.venv/bin/meural-mcp.
For a system install:
sudo useradd --system --home /var/lib/meural-mcp --create-home --shell /usr/sbin/nologin meural-mcp
sudo install -d -o meural-mcp -g meural-mcp /var/lib/meural-mcp
sudo cp examples/systemd/meural-mcp-*.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable --now meural-mcp-daemon.service meural-mcp-api.service
For a user install:
mkdir -p ~/.config/systemd/user
cp examples/systemd/meural-mcp-user.service ~/.config/systemd/user/meural-mcp.service
cp examples/systemd/meural-mcp-api-user.service ~/.config/systemd/user/meural-mcp-api.service
systemctl --user daemon-reload
systemctl --user enable --now meural-mcp.service meural-mcp-api.service
The user service samples expect meural-mcp at ~/.local/bin/meural-mcp.
Adjust ExecStart if you installed into a project virtual environment instead.
For LAN access, keep meural-mcp-api.service on 127.0.0.1:8733 and expose it
through HTTPS with Caddy or nginx. The shared token is still required by
MeuralMCP, but TLS should protect the token in transit.
The API service also exposes a remote streamable-HTTP MCP endpoint at /mcp/.
When the API is behind HTTPS, coding agents that support remote MCP should
connect directly to that URL with the same bearer token used by the REST API.
Clients do not need the MeuralMCP package installed just to use this hosted MCP
endpoint.
Set MEURAL_MCP_ALLOWED_HOSTS on the API service to the HTTPS hostnames clients
will use for remote MCP, such as meural-mcp.example.test or
meural-mcp.example.test:443. This keeps the MCP transport's DNS-rebinding
protection enabled while allowing your reverse-proxy hostname.
Example remote MCP URL:
https://meural-mcp.example.test/mcp/
Run the stdio MCP adapter only for coding agents that do not support remote MCP:
meural-mcp --storage-dir ~/.config/meural-mcp mcp --transport stdio
If this compatibility adapter runs on your workstation while the daemon/API run
on another host, point it at the remote API instead. In this mode,
set_device_image reads the image path from the workstation and uploads the
bytes to MeuralMCP:
export MEURAL_MCP_API_URL="https://meural-mcp.example.test"
export MEURAL_MCP_API_TOKEN="..."
export MEURAL_MCP_API_VERIFY_TLS=false # only for self-signed lab certs
meural-mcp mcp --transport stdio
MCP tools include:
list_devicesget_device_statusget_device_imageset_device_imageset_device_image_data
list_devices and get_device_status return the configured device name,
display name, cloud ID, LAN IP, orientation, enabled flag, reachability, current
assigned image, and the last recorded state for each device.
For hosted remote MCP, prefer set_device_image_data, which accepts base64 image
bytes plus a filename. set_device_image accepts a path that is local to the
machine running the MCP server, so it is mainly useful for stdio adapters or
server-local files.
Both image-writing tools store the image only after validating orientation and successfully loading the preview to the device. They return a failed result if the image cannot be parsed/thumbnailed, the orientation is wrong, the device is unknown, the file is missing, or the preview write fails.
Upload an image:
curl -X PUT http://127.0.0.1:8733/devices/canvas-1/image \
-H "Authorization: Bearer change-me" \
-H "Content-Type: image/jpeg" \
--data-binary @image.jpg
Cloud Init Details
Cloud init requires a Meural username/password.
meural-mcp --storage-dir ~/.config/meural-mcp init-cloud \
--username "$MEURAL_USERNAME" \
--password "$MEURAL_PASSWORD"
On a Linux host with user systemd available, add --install-systemd to write,
enable, and start meural-mcp.service. Interactive runs prompt for this unless
--no-systemd-prompt is supplied.
init-cloud:
- Authenticates to Meural cloud with the supplied username/password.
- Discovers cloud device IDs, names, LAN IPs, and writes them to
config.json. - Generates a shared API token, writes it to
config.json, and prints it so you can save it for REST/MCP clients. - If
config.jsonalready exists, writes a timestamped backup first. - Finds or creates landscape and/or portrait blank galleries, depending on the discovered devices.
- Ensures each blank gallery has a single correctly oriented blank image if the gallery is empty.
- Assigns each configured device to the matching blank gallery for its orientation.
- Sets
imageDuration=86400,previewDuration=86400,overlayDuration=120. - Syncs configured devices.
After init, review and edit config.json if you need to adjust device names,
cloud IDs, LAN IPs, orientations, or enabled flags before running the daemon.
FAQ
Why does this poll the devices?
Polling is not elegant, but it is deliberate. Meural frames can lose the loaded preview after sleep, boot, or other internal resets, so the daemon checks periodically and reloads the assigned preview when needed.
Why does setup use a blank image/gallery?
The blank image is a holding pattern. It keeps the device on a single neutral gallery by default, which makes the transition to locally managed previews a little less awkward than leaving whatever cloud playlist happened to be active.
Can I block the frames from the internet after setup?
Probably, but this still needs testing. Once cloud init has discovered devices, set the timeout values, and assigned the blank galleries, users may be able to firewall the Meural devices from the internet so future Meural cloud changes cannot alter or disturb them.
What would make this better?
A better solution likely needs hardware-level investigation. A community
teardown/hacking post describes a Meural RK3288 board and locating G, R, and
T test pads for ground, RX, and TX serial console access. Someone needs to
open a device, attach to those serial pins, and see whether there is a cleaner
local-control path than periodic preview reloads.
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.