// SCAN REPORT

Visual Annotation MCP

36 tools. 24 can modify or destroy data without limits.

24 write tools that can modify data. Rate limits recommended.

Last updated: 3 Jul 2026

24 can modify or destroy data

12 read-only

36 tools total

Community server · catalogue entry verified 03/07/2026

How to control Visual Annotation MCP ↓

TOOL MAP

What Visual Annotation MCP exposes to your agents

Read (12) Write / Execute (24) Destructive / Financial (0)

Running more than one server? Check your whole stack →

MOST DANGEROUS TOOLS

High Risk

The most dangerous Visual Annotation MCP tools

Execute · High navigate This tool executes a browser navigation action to an arbitrary URL supplied by the AI agent. Execute · High run_flow The tool executes programmatic flows rather than simply reading or passively inspecting. Execute · High run_flow_v2 A flow-running tool on a browser automation server would execute a sequence of actions whose effects depend on the flow definition and parameters provided. Execute · Medium wait_for_selector This tool executes a wait/polling operation that monitors DOM state and can block execution flow until a condition is met.

24 of Visual Annotation MCP's 36 tools can modify, destroy, or commit something on every call — and an agent calls them with no built-in limits.

POLICY

How to control Visual Annotation MCP

PolicyLayer is an MCP gateway — it sits between your AI agents and Visual Annotation MCP, and nothing reaches the server without passing your rules. These are the rules we recommend:

Rate limit write operations

{
  "annotate_last_image": {
    "limits": [
      {
        "counter": "annotate_last_image_per_hour",
        "window": "hour",
        "max": 30,
        "scope": "grant"
      }
    ]
  }
}

Prevents bulk unintended modifications from agents caught in loops.

Cap read operations

{
  "assert_element_exists": {
    "limits": [
      {
        "counter": "assert_element_exists_per_minute",
        "window": "minute",
        "max": 60,
        "scope": "grant"
      }
    ]
  }
}

Controls API costs and prevents retry loops from exhausting upstream rate limits.

Create a free account and register Visual Annotation MCP — nothing to install.
Add these rules — paste them, or build them visually. Tune the limits to your setup.
Point your MCP client (Claude, Cursor, anything) at your gateway URL.

ENFORCE POLICY ON VISUAL ANNOTATION →

Instant setup, no code required.

FULL CATALOGUE

All 36 Visual Annotation MCP tools

EXECUTE 18 tools

Execute check_uncheck Check or uncheck a checkbox/radio using selector or element id. Execute click_by_role Click by accessible role and optional name. Execute click_by_selector Click the first element matching a CSS selector. Execute click_by_text Click the first actionable interactive element whose text/label matches. Execute click_element Click an inspected element id, waiting for it to become actionable first. Execute close_cookie_banner Best-effort cookie banner dismissal helper. Execute dismiss_common_popups Best-effort dismiss of cookie/modals using common labels and close icons. Execute dismiss_overlay Dismiss an overlay via selector or auto heuristics (+ optional Escape). Execute highlight_element highlight_element Execute navigate Go to a URL (HTTP/S). Clears element ids from a previous inspect. Execute press_key Press a keyboard key (for example: Enter, Escape, Tab). Execute run_flow Execute a multi-step interaction flow (legacy alias). Execute run_flow_v2 run_flow_v2 Execute select_option Select option on a <select> by value/label/index using selector or element id. Execute submit_form Submit a form by form selector/id or clicking submit-capable element. Execute wait_for_selector Wait for a selector to reach attached/detached/visible/hidden state. Execute wait_for_text Wait for text to become visible globally or under a selector scope. Execute wait_for_url Wait until current URL contains the given fragment (useful between flow steps).

WRITE 6 tools

Write annotate_last_image annotate_last_image Write assert_element_visible Assert that an element is visible/actionable. Write fill_by_label Fill a field by its associated label text. Write fill_by_selector Fill a field selected by CSS selector (no inspect id needed). Write fill_element Fill an inspected input-like element id with text. Write upload_file Upload a local file into a file input using selector or element id.

READ 12 tools

Read assert_element_exists Assert that an element exists by selector or inspected element id. Read assert_text_contains Assert that page (or selector scope) contains specific text. Read assert_url_matches Assert that current URL contains a pattern or matches regex. Read detect_blockers Detect likely viewport blockers (modals/overlays) and return JSON metadata. Read extract_element Extract structured data from one element (tag/id/role/text/attrs/box). Read extract_form_data Extract key/value form field data from a form selector. Read extract_page_model Extract a lightweight page model (headings, landmarks, forms, interactive count). Read extract_table Extract table headers and row objects from a table selector. Read inspect_elements Return JSON of visible interactive/form controls with ids (e0, e1, …). Read observability_snapshot Return in-memory per-tool latency/failure metrics snapshot as JSON. Read screenshot_element Screenshot a single DOM element by id from the last inspect_elements (tight crop). Read screenshot_viewport Capture the current page (viewport or full scrolling page).

EXPLORE

Related servers

Other MCP servers with similar tools — same risk classification, starter policies for each.

FAQ

Questions about Visual Annotation MCP

How do I prevent bulk modifications through Visual Annotation MCP? +

The Visual Annotation MCP server has 6 write tools including annotate_last_image, assert_element_visible, fill_by_label. Set a rate limit in your policy -- for example, 10 calls per hour prevents an agent from making more than 10 modifications per hour. PolicyLayer enforces this at the gateway, before calls reach Visual Annotation MCP.

How many tools does the Visual Annotation MCP server expose? +

36 tools across 3 categories: Execute, Read, Write. 12 are read-only. 24 can modify, create, or delete data.

How do I enforce a policy on Visual Annotation MCP? +

Register the Visual Annotation MCP server in PolicyLayer, apply the suggested rules above (adjust the limits to your use case), and point your AI client at the PolicyLayer proxy URL instead of the server directly. Your agents keep the same tools; PolicyLayer evaluates every call against policy before it executes. Nothing to install, live in minutes.

Enforce policy on every Visual Annotation MCP tool call.

Deterministic rules across all 36 Visual Annotation MCP tools. Per-identity grants. Full audit log. Live in minutes. Nothing to install.

ENFORCE POLICY ON VISUAL ANNOTATION →

Instant setup, no code required.

36 Visual Annotation MCP tools catalogued and risk-classified — across an index of 43,000+ MCP servers.