🤖 Computational Research

A Wolfram-centric Claude plugin for AI-assisted computational research.

• 📁 Turns a folder with resources (code, PDFs, Markdown, notebooks, ...) into an organized git repo, which it maintains.
• 🐺 Imports and exports Wolfram notebooks via conversion to Markdown.
• 📚 Grows and maintains a wiki knowledge base.
• 🔍 Gathers and summarizes resources, keeping a Markdown summary and recovery options (e.g. where to download).
• 📦 Converts code into a paclet and helps with its build and deployment (e.g. with or without docs, local build or cloud).
• 📓 Generates expository Wolfram notebooks and helps publish them on Wolfram Cloud (explicitly marked as LLM-generated, keeping a Markdown mirror).
• 📝 Scaffolds a LaTeX or Typst paper and edits the user-owned document on request.
• 🧬 Optionally records the prompt and intent behind every generated artifact, keeping it traceable.
• 📔 Optionally keeps a running scientific journal in LaTeX or Typst — concise, cited def/thm/rem entries of what was learned.
• 🧭 Offers a guided tour through the project for a human, and a revision protocol (stateful for multi-session work).
• ✅ Tracks plans, todos, and state.

⚠️ Disclaimer. This repo grows on the fly out of my own thoughts and needs around AI assistance in computational research. It is a working draft, in need of human revision and selective improvement. Helpers and testers welcome!

📥 Installation

The plugin is distributed through the WolframInstitute plugin marketplace.

Claude Code (CLI / VS Code extension) — the author's setup:

claude plugin marketplace add WolframInstitute/ClaudePluginMarketplace
claude plugin install computational-research@WolframInstitute

Claude Desktop app — install from the marketplace GUI.

Note: Operation in Cowork mode and Chat mode has not been tested.

⚙️ Recommended Setup

The plugin works best with Wolfram Engine (or Mathematica).

MCP servers — the plugin draws on these:

Server	Required	Purpose	Source
Wolfram (official)	yes	Evaluation, notebook I/O, docs search, tests	Wolfram/AgentTools
arxiv-latex-mcp	recommended	Download LaTeX source of arXiv papers	takashiishida/arxiv-latex-mcp
arxiv	recommended	Search and download arXiv papers	blazickjp/arxiv-mcp-server
wolfram (unofficial)	optional	Additionally supports Wolfram Language LSP, similar to Serena	sw1sh/WolframMCP

Installing the official Wolfram server:

InstallMCPServer["ClaudeCode", "WolframLanguage"]

Note: On older Wolfram versions, the legacy Wolfram/MCPServer paclet still works as a fallback if that is what you have installed.

🔑 License seats. Each running kernel — every Wolfram MCP server, every open front-end, every wolframscript call — consumes one of your license's $MaxLicenseProcesses seats. The plugin is MCP-first: it routes Wolfram work through the official server's single persistent kernel and treats the .wls scripts as a fallback for when no MCP is attached (e.g. headless runs). Before spawning wolframscript it checks headroom and warns rather than failing opaquely. Running both Wolfram MCP servers at once uses two seats — /check-env reports live headroom and flags this.

🧩 Skills

Skills share their name with the matching slash command (one name per feature), grouped by domain.

Skill	Description
new-project	Scaffold a new project (research, math, paclet-dev, paclet)
scaffold-paper	Scaffold a LaTeX or Typst paper, then edit it on request
journal	Keep an optional cited LaTeX/Typst journal (def/thm/rem), off by default
init-wiki	Create a markdown knowledge base (Wiki/)
update-wiki	Update wiki articles, index, and log
check-wiki	Audit the wiki for staleness and gaps
search-wolfram	Search Wolfram docs, Function Repository, Community, writings
search-math	Search MathWorld, nLab, OEIS, DLMF, Wikipedia math
add-resource	Add a paper, repo, or page with recovery info
cite	BibTeX from an arXiv ID or DOI
new-notebook	Build Wolfram notebooks from Markdown
lean	Drive a Lean/Mathlib formalization session
build-paclet	Build a paclet and install it locally
publish-paclet	Build, install, and publish a paclet to the Cloud
work	Manage multi-session work items (spec, tasks, progress)
next-session	Run the next task in a fresh session, then stop
provenance	Track the prompt behind each generated artifact
start-tour	Run a guided tour of the project
revise	Human revision protocol for deliverables (no command)

⌨️ Slash Commands

Command	Description
/computational-research:new-project	Scaffold a new project
/computational-research:scaffold-paper	Scaffold a LaTeX or Typst paper
/computational-research:journal	Toggle, scaffold, add, or list journal entries
/computational-research:init-wiki	Create the Wiki/ knowledge base
/computational-research:update-wiki	Update the wiki after changes
/computational-research:check-wiki	Audit the wiki
/computational-research:search-wolfram	Search the Wolfram ecosystem
/computational-research:search-math	Search external math resources
/computational-research:add-resource	Add a resource to the wiki
/computational-research:cite	BibTeX from an arXiv ID or DOI
/computational-research:new-notebook	Create a Wolfram notebook
/computational-research:lean	Drive a Lean/Mathlib session
/computational-research:build-paclet	Build and install a paclet
/computational-research:publish-paclet	Publish a paclet to the Cloud
/computational-research:work	Create or manage a work item
/computational-research:next-session	Run the next task, then stop
/computational-research:provenance	Toggle/show prompt tracking
/computational-research:start-tour	Start or resume a guided tour
/computational-research:check-env	Check kernel and MCP availability
/computational-research:load-project	Summarize project status

📄 License

MIT