도구

Two-stage deep research: drafts an editable outline for the topic, then parallel subagents search the web item by item and compile a Markdown report. Good for literature reviews, competitor research and due diligence. Install by copying files per the README, plus pip install pyyaml.

Apify's official skill collection: picks from 130+ curated Actors to scrape public data from social networks, maps, e-commerce sites and more, and can also build, package and deploy Actors. Needs an Apify account; a free tier is available.

Four tools for real-time web search, page content extraction, site mapping and full-site crawling, available as a remote server or run locally. Maintained by Tavily; needs an API key or OAuth sign-in, and a free account is available.

Turns a paper's method description (text or PDF) or data files into publication-ready diagrams and plots, with several AI reviewer roles refining each figure. Needs an OpenAI or Gemini API key (billed per image), Python and uv. A community reimplementation, not by the original paper's authors.

Translates entire books: splits PDF, DOCX or EPUB into chunks for parallel subagents, keeps terminology consistent, resumes interrupted runs and verifies with hashes, then outputs HTML, DOCX, EPUB or PDF. Needs Calibre, Pandoc and Python.

MiniMax's official CLI, built for agents: chat with MiniMax models from the terminal and generate images, video and speech, plus speech transcription, image understanding and web search. Works with the international platform. Requires a MiniMax Token Plan subscription.

A CLI and Python library by Simon Willison: imports JSON and CSV straight into SQLite with automatic table creation, runs in-memory SQL queries over files, and supports full-text search and table transformations.

Unofficial Intercom CLI built with Printing Press, covering the full REST API: sync conversations and contacts for offline search, bulk-tag conversations about one incident, see customer overviews and SLA reports, and pull or push help articles. US, EU and AU regions; needs an access token.

The Linux Foundation's open-source compliance toolchain: analyzes project dependencies, scans licenses and copyrights, checks known vulnerabilities, flags policy violations and generates SBOMs and notice files. Needs Java 25 or Docker and ideally 8 GB RAM; have legal review license findings.

Self-hostable long-term memory for agents, offering MCP, a REST API and a web dashboard: semantic search, knowledge graph and automatic memory consolidation. Embeddings are computed locally; choose SQLite, Cloudflare or Milvus as the backend.

550+ medical research skills from AIPOCH: PubMed search strategy, study design and sample size, statistics and bioinformatics, journal writing and reviewer responses, all vetted by MedSkillAudit. Some need Python/R or database credentials; have researchers and clinicians review results.

Fetches US, Hong Kong and China A-share stock data with zero runtime dependencies: real-time quotes, daily, weekly and monthly candles and search as JSON, falling back across Tencent, Sina and Eastmoney sources. Also runs as an MCP server. Unofficial public APIs, no stability guarantee.

Before tool calls run, blocks destructive Git and file commands (git reset --hard, rm -rf) and access to sensitive files like SSH keys and .env. Parses command semantics, so bash -c wrappers are caught too. Works with Claude Code, Codex and other coding CLIs; needs Node.js 18+.

A three-stage workflow for math modeling contests like MCM/ICM and CUMCM: analysis, solving and plotting in Python or MATLAB, and a Word (optionally LaTeX) paper, with stage-check subagents and OpenAlex plus AnySearch literature search. Generated papers are for reference only.

Multi-engine web search MCP with no API key, supporting Bing, DuckDuckGo, Brave, Baidu, Sogou, CSDN, Juejin and more; also fetches page content, CSDN articles and GitHub READMEs. Ships a CLI and a local server too.

MathWorks' official server that lets agents start your local MATLAB, write and run scripts and unit tests, and check code style and potential errors, for computing, simulation and data analysis. Needs a licensed MATLAB R2021a+ and the downloaded server binary.

Command-line reference and bibliography manager: add papers by DOI with metadata fetched automatically, search by author, title or tag, and import or export BibTeX. Each entry is a readable YAML file that agents can read and write easily; a web UI is included. Needs Python.

Recommends daily papers from arXiv and Semantic Scholar by research interest, reads them and writes illustrated Obsidian notes and a paper graph; can also search venues like CVPR and NeurIPS. Needs Python 3, Obsidian, a vault path and an interests config; the web app needs an Anthropic API key.

From Harvard Medical School's Zitnik Lab: one MCP connecting 1,000+ scientific tools and databases for drugs and targets, genes and proteins, clinical trials, FDA data and PubMed, for biomedical and drug research. Runs with uv; some sources need API keys; medical conclusions need expert review.

Lets agents read and write your local Obsidian vault directly: search notes, append to daily notes, tidy tags and properties, and summarize reading or research notes by topic. No need to open Obsidian or install plugins; access is limited to the vault and can be read-only. Needs Node.js 20+.

Econometrician Pedro Sant'Anna's academic Claude Code setup (fork the repo to use it): Beamer/Quarto lecture slides, paper review and mock refereeing, R/Python analysis and replication packages, with specialist subagents and quality gates. Needs Git and Python; slides need XeLaTeX and Quarto.

Lets agents drive WinDbg's CDB/KD to analyze crash dumps, debug user-mode processes remotely and debug kernels, and triage many dumps in a batch. Windows only; requires WinDbg or Debugging Tools for Windows.

Reads and writes Obsidian vaults from the terminal: search, view, create, edit, move and delete notes, and manage frontmatter and daily notes, without Obsidian running. Good for agents maintaining a personal knowledge base. Formerly Obsidian CLI.

Web search through DuckDuckGo plus fetching and parsing page content, with built-in rate limiting and LLM-friendly result formatting. No API key needed; runs with a single uvx command.