Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    tooldrop

    1

    Turn any website task into a fast, reusable MCP tool with local execution and auto-repair capabilities.

    Free

    2 usesSecurity scanned
    tooldrop

    tooldrop

    Example session with this skill installed

    Check if there are any existing tools for amazon.com to list my recent orders, and if not, build one that returns order numbers and delivery dates.

    • Read your context and instructions
    • Compiled the tooldrop

    I have checked for existing Amazon tools and found no direct match for your specific order history task. I will now initiate the build process for a new tool.

    Build Job: Amazon Order History

    PropertyValue
    Tool Nameamazon_list_orders
    Statuswaiting_for_answer
    TaskList recent orders with order numbers and delivery dates
    Authorityread
    Job IDjob_8829_amzn_read

    The tool builder is currently at the login screen. It has detected that a sign-in is required to access your order history.

    Action Required:
    Please provide the 6-digit 2FA code sent to your mobile device or email so I can proceed with the tool generation.

    Next steps

    1. Send the 2FA code in the chat so I can call answer_question.
    2. Wait approximately 60 seconds for the build to complete after authentication.
    3. Test the tool with the sample output I will provide once the status reaches succeeded.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Stop manually scraping dashboards or driving browsers for repetitive web tasks. This skill turns complex website actions into local, typed MCP tools that execute in seconds. It eliminates the latency of model-based navigation by generating plain code to handle data extraction and form submissions.

    What it does

    • Generate tools by converting natural language tasks into executable code for any website.
    • Automate extraction of structured data from dashboards, invoices, listings, and orders.
    • Handle authentication via encrypted local storage and managed 2FA workflows.
    • Execute writes safely using a human-in-the-loop approval system with visual previews.
    • Monitor changes by scheduling watches that POST data to webhooks when rows change.
    • Auto-repair tools when website layouts shift, ensuring long-term reliability without manual updates.

    How it works

    1. Search for existing tools using find_tool or browse proposals with discover_tools.
    2. Build new tools by providing a URL and a plain-text task description to build_tool.
    3. Configure authentication by having the user run local login commands for sensitive sites.
    4. Execute the resulting MCP tool directly with typed inputs for fast, reliable results.

    Frameworks & tools

    Built on Node.js 24+, Playwright for browser automation, and Ollama for local model execution. It operates as a standard MCP server compatible with Claude Code, Claude Desktop, and Cursor.

    Why this beats prompting it yourself

    Standard agents struggle with session persistence, 2FA, and the high token cost of repeated browser navigation. This skill moves the logic to local code, making subsequent calls nearly instant and significantly cheaper while adding robust error handling for site changes.

    Use cases

    • Inventory tracking by pulling live stock levels from supplier dashboards hourly.
    • Financial reporting by gathering invoices from SaaS portals into structured JSON.
    • CRM updates by building write tools to sync lead data into internal web forms.
    • Price monitoring by watching competitor pages and triggering webhooks on changes.

    Known limitations

    Requires Node.js 24 and a local Ollama installation for the build phase. Cannot bypass CAPTCHAs or paywalls; these require manual user intervention via the human_needed flow.

    Requires Node.js 24+ and Ollama. Cannot bypass CAPTCHAs or paywalls. Write tools require an initial manual recording via CLI or extension.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    2 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 16 days ago

    • Free with an Agensi account

    Requires

    Node.js

    Permissions

    Browser
    Network Access
    Listed16 days ago

    What's inside

    Frequently Asked Questions

    While standard agents navigate sites step-by-step, tooldrop converts tasks into local code tools. This eliminates high token costs and latency, providing near-instant execution for repetitive tasks like data extraction or form filling.

    tooldrop is a standard Model Context Protocol (MCP) server, making it fully compatible with Claude Desktop, Claude Code, Cursor, and any other IDE or agent that supports the MCP standard.

    Yes, tooldrop includes an auto-repair feature that detects when a website's layout or DOM structure has changed. It automatically regenerates the underlying code to ensure your tools keep working without manual intervention.

    tooldrop handles 2FA and sensitive logins by utilizing encrypted local storage and a 'human-in-the-loop' workflow. Users perform the initial login locally, allowing the tool to securely manage sessions for subsequent automated runs.

    You will need Node.js 24+ and a local installation of Ollama to handle the initial tool generation phase. The browser automation itself is powered by Playwright, which is bundled with the skill.

    For security, all 'write' actions (like submitting forms or deleting data) trigger a visual preview and require explicit user approval before the action is executed on the live site.

    New in Agents & Orchestration