Command-Line Interface¶

CellPainting-Claw exposes two public CLI surfaces:

cellpainting-skills
cellpainting-claw

For most users, the right starting point is cellpainting-skills.

Choosing A CLI¶

User goal	Start with	Rationale
run one clear named task such as segmentation, pycytominer, or DeepProfiler	`cellpainting-skills`	this is the shortest path from a user goal to a documented task
inspect configuration, use data-access helpers, or expose MCP tools	`cellpainting-claw`	this is the direct toolkit CLI

`cellpainting-skills`¶

cellpainting-skills is the named-task CLI.

Its three main commands are:

Command	Role
`list`	show the public skill catalog
`describe`	show what one skill does
`run`	execute one named skill

First Session¶

CONFIG=configs/project_config.demo.json

cellpainting-skills list
cellpainting-skills describe --skill cp-extract-segmentation-artifacts
cellpainting-skills run \
  --config "$CONFIG" \
  --skill cp-extract-segmentation-artifacts \
  --output-dir outputs/demo_segmentation

This is the main CLI path that new users should understand first:

inspect the available skills
inspect one skill
run one named task

The public CLI skill catalog currently covers:

data access inspection, planning, and download
CellProfiler measurement extraction and single-cell table construction
pycytominer aggregation, annotation, normalization, feature selection, and summarization
segmentation preparation, artifact export, preview generation, and crop export
DeepProfiler input export, project building, model execution, feature collection, and summarization

`cellpainting-claw`¶

Use cellpainting-claw when you need commands that are more specific than a named skill.

The most useful command groups for normal users are:

Command group	Scope	Example commands
configuration inspection	inspect config state and other advanced runtime selections	`show-config`, `show-cppipe-selection`, `validate-cppipe-config`
data access	inspect data sources, build plans, and execute downloads	`summarize-data-access`, `plan-data-access`, `execute-download-plan`
direct toolkit runs	run toolkit-side commands when you want lower-level control than a public skill	`run-segmentation`, `run-profiling`, `run-deepprofiler-pipeline`
MCP bridge	expose or inspect the MCP interface used by agent runtimes	`serve-mcp`, `list-mcp-tools`, `run-mcp-tool`

Data Access Commands¶

CONFIG=configs/project_config.demo.json

cellpainting-claw summarize-data-access \
  --config "$CONFIG"

cellpainting-claw plan-data-access \
  --config "$CONFIG"

MCP Bridge¶

cellpainting-claw serve-mcp --transport stdio

This is the bridge interface used by OpenClaw and other MCP-capable clients.

Lower-Level Commands¶

The toolkit CLI also contains additional lower-level or compatibility commands. Those remain available, but they are not the main starting point for most users, so this page focuses on the command paths that match the public skill model.

This is also where advanced inspection commands remain available when needed.