도구

Lets coding agents read your codebase, extract a design system, then generate UI mockups, slide decks, posters and other marketing visuals on Superdesign's infinite canvas and iterate in branches. Needs @superdesign/cli and a Superdesign account; free to start.

Sentry's official remote MCP: lets coding agents query errors, issues, stack traces and performance traces, and call Seer for root-cause analysis. Sign in with OAuth, or connect to self-hosted Sentry over stdio.

Lets agents manage Google Cloud resources in natural language through the gcloud CLI, blocking interactive and other agent-unfriendly commands by default, with configurable allow and deny lists. A preview release from Google, not an officially supported product; install gcloud and sign in first.

Companion skill to the User Interface Wiki: 152 rules in 12 categories, from motion, easing and exit animations to UI sounds, interaction laws, prefetching and typography, for writing or reviewing web UI code, with issues listed by file:line. Needs an existing web project and Node.js to install.

AWS Labs' official plugin marketplace: deploy-on-aws (architecture recommendations, cost estimates and IaC deployment), aws-serverless, aws-amplify, databases-on-aws, sagemaker-ai and more. AWS suggests Agent Toolkit for AWS for new projects.

HashiCorp's official Terraform and Packer skills: write HCL in the official style, refactor modules, terraform test, provider development, Stacks and policy, plus AWS and Azure image builds.

HashiCorp's official plugin marketplace: the terraform plugin has 16 skills (configuration, modules, provider development, testing, import, Stacks and policy), and the packer plugin has 4 image-build and HCP Packer skills.

Answers macro questions with market data: combines 13 free sources including Polymarket and Kalshi prediction markets, US Treasury yields, CFTC positioning, SEC insider trades and A-share fund flows into probability estimates, reasoning and scenarios. No API key; needs uv. Not investment advice.

Jina AI's official remote MCP: converts web pages and PDFs to Markdown, searches the web, arXiv, SSRN and images, and offers screenshots, reranking and semantic dedup. Reading tools work without a key (rate-limited); search and rerank need a Jina API key.

A monthly CFO routine for small companies and agencies: pulls bank, receivables, payroll and expense data, categorizes and reconciles it, computes month-end cash, updates forecasts and writes a management snapshot, plus a weekly cash check. Needs your own config and account data; review the results.

Runs SQL queries against ClickHouse and browses databases and table schemas, read-only by default; can also enable embedded chDB to query data directly in files and URLs. Maintained by ClickHouse.

Drives desktop Excel on Windows: refresh Power Query, build pivot tables and charts, evaluate DAX and run VBA while keeping existing formulas and formatting, so finance and admin staff can automate existing reports. Needs Windows, Excel 2016+ and Node.js; no macOS.

Helps Indian resident individuals prepare income tax returns: the agent reads Form 16, AIS and more, asks about deductions, computes tax with a tested Python program, compares old and new regimes and builds a filing checklist; you pay and file yourself. AY 2026-27 only; Python 3.

Intlayer's official MCP server: in projects using Intlayer, lets agents build dictionaries, fill in missing translations with AI, sync multilingual content with the Intlayer CMS and look up docs. Needs Node.js; AI translation needs your own LLM API key, CMS sync needs credentials.

From Vercel: discovers locally running Next.js 16+ dev servers and hands the runtime errors, routes and logs from their built-in MCP endpoint to the agent, and points the agent to docs matching your installed version.

Unofficial Yahoo Finance data tool with a yahoo-finance CLI, an MCP server and an agent skill: stock quotes, price history, financials, option chains and analyst ratings. Needs Node.js 22+, no API key. The API is unofficial and may change; for research only.

Unbabel's machine-translation quality metric: comet-score rates translations per sentence and overall, XCOMET marks error spans and severity, and CometKiwi works without a reference. Good for spot-checking MT. Needs Python 3.8+; models download from Hugging Face, some need a license; GPU helps.

Enter a URL and get a single-page SEO audit report in HTML: checks title and meta description, H1, robots, sitemap, structured data and more, and lists priority fixes. Needs local Python; the full audit also needs a Google PageSpeed API key.

Interviews you on scope, pain points and brand tone, infers the visual language from existing styles, and builds five React UI variants on a temporary route with a feedback overlay. Iterates on feedback, then writes an implementation plan and deletes the temp files.

Shopify's official CLI: scaffold, develop locally and deploy Shopify apps, extensions, Functions and themes, and build and host Hydrogen custom storefronts. Requires Node.js.

dbt Labs' official plugin marketplace with dbt and dbt-migration plugins: write models and unit tests, build semantic-layer metrics, govern with dbt Mesh, debug failed jobs, and handle version upgrades and cross-platform migrations. Most skills need an existing dbt project.

dbt Labs' official skills: write and edit dbt models, add unit tests and docs, set up the MetricFlow semantic layer, debug failed jobs, and handle dbt version upgrades and cross-platform migrations. Most skills need dbt installed and an existing project.

US stock research: /trading-ideas:research gathers earnings, peer valuations, unusual options and insider trades via web search and writes an analyst-style report with rating, price target and bull/bear cases. Needs a paid Claude plan. Web data can be wrong; not investment advice.

A research plugin for Claude Code: commands like /omp:survey, /omp:experiment and /omp:write chain literature review, topic selection, experiments, writing and self-review, with 5 specialized agents and 34 research skills that remember project progress across sessions. Experiments lean toward ML.