Skip to main content
Glama

desktop_watch_start

Start a live desktop observation session that captures changed screen frames and OCRs them on demand. Use it to monitor screen activity while keeping content in memory only.

Instructions

Start a LIVE desktop observation session (explicit user request only).

The screen is sampled every interval seconds and OCR'd only when the frame actually changed, so an idle screen costs almost nothing. OCR runs off the MCP event loop. Screen content is returned on request and NEVER written to disk, the audit log or memory. Poll desktop_watch_read() for the latest observation (MCP has no push channel) - the console shows the same events.

ocr=False watch the screen as IMAGES instead of text: frames are still captured and compared, but nothing is recognised. Use it when no OCR engine is installed and a vision model will look at the frames (desktop_watch_read(as_image=True)).

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "desktop_watch_start"
  }
}

Input Schema

NameRequiredDescriptionDefault
ocrNo
regionNo
monitorNo
intervalNo
languageNoeng
providerNo

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv0.4.0

TDQS

A4.4/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations exist, so the description carries the full burden and does so: sampling cadence, OCR-only-on-frame-change cost model, off-event-loop OCR, privacy guarantee that content is never written to disk/audit log/memory, and the explicit 'no push channel, you must poll' constraint. These are exactly the non-obvious traits an agent needs before starting a surveillance-style session.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded imperative sentence, short paragraphs, and the ocr=False block is set off visually. Slightly verbose in the polling/console sentence, but no line is filler.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Strong on runtime and privacy behavior for a zero-annotation tool with no output schema, but with 6 parameters at 0% schema coverage the description leaves half the surface (region format, monitor indexing, language/provider meaning) undocumented in either place.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 0%, so the description must explain six parameters and it explains only `interval` and `ocr` (the latter with a concrete conditional behavior and the paired read flag). `region`, `monitor`, `language`, and `provider` are left entirely to bare schema titles, which is a real gap, though the two explained parameters are explained well.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Names a specific verb and resource ('Start a LIVE desktop observation session') and immediately frames the scope (live, sampled, OCR-on-change), which separates it from siblings like desktop_inspect or desktop_click that act once rather than continuously.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives explicit preconditions ('explicit user request only') and names the follow-up sibling (poll desktop_watch_read()) plus the alternative mode for the no-OCR-engine case. It stops short of contrasting with desktop_inspect/eyes_see_desktop, so the agent must still infer when a one-shot inspection is preferable.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.