OpenHCS — Claude Code MCP server: what it does, how to install
MCP server to author, validate, and execute high-content microscopy analysis pipelines directly through AI coding agents.
claude mcp add openhcs -- uvx openhcsWhat OpenHCS does
OpenHCS unites visual desktop tools and Python scripts under an agent-ready pipeline engine. It relies on a compile-then-execute model. Before processing touches raw image sets, the typed compiler validates input data sources, step artifacts, memory contracts, and executor targets. Late failures stop before processing starts.
An agent connected via MCP inspects available function catalogs, builds typed configurations, and triggers processing runs. Output images, object masks, and measurements stream to PolyStore storage or live viewers like Napari and Fiji over an isolated ZMQ transport. Running the server requires Python 3.11 to 3.13 and the uvx command-line tool.
Pipelines remain bidirectional. You can adjust them visually, export them as Python code, or import existing CellProfiler .cppipe setups. Dedicated GPU hardware is not mandatory, since OpenHCS falls back to CPU concurrency via ProcessPoolExecutor.
What it helps with
- validate high-content microscopy pipelines before heavy batch execution
- convert legacy CellProfiler
.cppipefiles into executable Python workflows - stream intermediate assay images to Napari or Fiji in real time
- dispatch multi-well image processing across CPU workers or GPUs
Example prompts for the agent
- Inspect the microscopy images in this directory and build a typed cell analysis pipeline
- Convert
assay_pipeline.cppipeinto OpenHCS Python code and check its memory contracts - Run the compiled pipeline on the multi-well plate and stream output frames to Napari
How to install OpenHCS
Platform CLI
claude mcp add openhcs -- uvx openhcsuvx runs the command named after the package; if the README names another one, use uvx --from openhcs <command>
Platform CLI
codex mcp add openhcs -- uvx openhcsuvx runs the command named after the package; if the README names another one, use uvx --from openhcs <command>
Config file
{
"mcpServers": {
"openhcs": {
"command": "uvx",
"args": [
"openhcs"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from openhcs <command>
Config file
{
"servers": {
"openhcs": {
"type": "stdio",
"command": "uvx",
"args": [
"openhcs"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from openhcs <command>
Platform CLI
gemini mcp add openhcs uvx openhcsuvx runs the command named after the package; if the README names another one, use uvx --from openhcs <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about OpenHCS
How to install OpenHCS in Claude Code?
Run claude mcp add openhcs -- uvx openhcs. This sets up the server via uvx, requiring an environment with Python 3.11 to 3.13.
What does the OpenHCS MCP server do?
It provides agent tools to inspect image sets, construct typed analysis pipelines, validate execution plans, and run processing across wells or plates.
Is a dedicated GPU required to run OpenHCS?
No, GPU acceleration remains strictly optional. When no GPU is configured, OpenHCS runs bounded worker lanes on the CPU using ProcessPoolExecutor.
