# OpenHCS MCP server

> Lets an AI client inspect, author, validate, and execute high-content microscopy imaging workflows using a compile-then-execute architecture.

- Repository: https://github.com/OpenHCSDev/OpenHCS
- Page: https://hostedmcp.org/servers/openhcsdev/openhcs
- Transport: stdio
- Stars: 7
- License: MIT
- Latest version: 0.8.6

## Install

### Claude Code

```sh
claude mcp add openhcs -- uvx openhcs
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "openhcs": {
      "command": "uvx",
      "args": [
        "openhcs"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "openhcs": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "openhcs"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"openhcs","type":"stdio","command":"uvx","args":["openhcs"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "openhcs": {
      "command": "uvx",
      "args": [
        "openhcs"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add openhcs -- uvx openhcs
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.openhcs]
command = "uvx"
args = ["openhcs"]
```

### Gemini CLI

```sh
gemini mcp add -s user openhcs uvx -- openhcs
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "openhcs": {
      "command": "uvx",
      "args": [
        "openhcs"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `OPENHCS_CPU_ONLY` | no | no |  |

## FAQ

### Can I use OpenHCS with CellProfiler?

Yes, you can open .cppipe files in the desktop application or import them from Python into OpenHCS pipeline declarations.

### How does the MCP server work with AI agents?

The local stdio MCP server allows clients like Claude Desktop or ChatGPT to inspect functions, build and validate pipelines, run them, and inspect results.

### Does OpenHCS support GPU acceleration?

Yes, it supports GPU acceleration through libraries like CuPy, PyTorch, JAX, and TensorFlow, with compiled callable contracts selecting framework-local GPU devices.
