Skip to content

Tools Reference

The farscry MCP server exposes two tools. Workflows call them directly; no CLI piping required.

farscry_extract

Converts a screenshot to VASP structured context.

{
"name": "farscry_extract",
"description": "Converts a screenshot into VASP structured context for automation tools",
"parameters": {
"image_path": {
"type": "string",
"description": "Absolute path to image file"
},
"lang": {
"type": "string",
"default": "eng",
"description": "Language code - e.g. eng, por, eng+por"
},
"affordances": {
"type": "boolean",
"default": true,
"description": "Include affordance list in output"
}
}
}

Returns: Full VASP output - screen_type, agent_context, ui_tree, affordances, state_id.

farscry_diff

Returns the semantic delta between two screenshots.

{
"name": "farscry_diff",
"description": "Returns semantic delta between two screenshots - appeared, changed, removed",
"parameters": {
"before": {
"type": "string",
"description": "Absolute path to before image"
},
"after": {
"type": "string",
"description": "Absolute path to after image"
}
}
}

Returns: VASP diff output : entries array of DeltaEntry items, each typed as appeared, changed, removed, or unchanged. Also includes context_similarity, context_changed, agent_context, and tokens_saved.

Workflow usage

With the MCP server running, workflows call the tools directly:

Workflow calls farscry_extract(image_path="/tmp/before.png")
-> receives full VASP context
Workflow takes action
Workflow calls farscry_diff(before="/tmp/before.png", after="/tmp/after.png")
-> receives typed delta of what changed

No piping, no subprocesses - the workflow uses the tools the same way it uses any MCP tool.