Open Research Workspaces Index
April 23, 2026 · View on GitHub
Public research notebooks — living repositories where an open question is explored iteratively with Claude Code as the execution engine. Each workspace pairs a research question with context, prompts, and accumulated outputs.
Last Updated: April 2026
Table of Contents
Research Workspaces
Claude ADHD Research Workspace
Living research notebook on ADHD drug policy and access — a real project open-sourced as both a working workspace and a worked example of the research-repo pattern. Question: how should ADHD medication access policy be structured, and what does the evidence actually say?
Claude Local AI Agent Research
Evolving workbook tracking agentic LLMs and CLI tools for desktop/server system administration (not code generation). Question: which local/agentic models and tools are actually suited to sysadmin workflows, and where is the gap between code-gen-centric tooling and filesystem-wide admin use?
GitHub Repo Org Question
Ongoing stack/tooling research notebook on managing a large personal GitHub repo fleet. Question: how do you manage hundreds of personal repos across multiple machines without losing time to navigation, sync accidents, or filesystem churn? Architecture currently settled on a five-layer stack keyed on git remote rather than filesystem paths.
Live Typing UX Research
Public research workspace cataloguing the UX of live voice typing as a desktop keyboard replacement. Question: what interaction patterns, friction points, and architectural tradeoffs (streaming vs offline ASR, partial vs final injection, endpointing) define a workable "ideal" desktop live-typing UI?
Local Claude Plugin Install Research
Research workspace investigating how to permanently install Claude Code plugins from a local filesystem path without requiring a public marketplace. Question: is there any mechanism — documented or undocumented — for persistent local plugin installation, and what is the most robust workaround until native support arrives?
MCP Server Research
Multi-model stack research run (January 2026) into MCP gateway solutions. Question: which MCP gateways best support dynamic tool discovery and mitigate context pollution for a solo developer? Compares Claude Opus, Perplexity Sonar, o4-mini, Gemini Flash, and Alibaba DeepResearch outputs.
Shared AI Chats Research
Open notebook exploring multi-user shared AI chat applications. Question: do collaborative AI chat tools exist that let couples/families/small teams share a single assistant in real time, and what would the UI, attribution, and context-window requirements look like?
State Of Claude Context 0426
Openly-logged Q&A research notebook on the current state of context handling in Claude as of April 2026. Question: what is the state of Claude's context window, prompt caching, memory, compaction, and tool-result handling — and how do these shape real-world workflows? Q&A pairs accumulate in questions/ + answers/, with periodic consolidation into PDFs.
Templates
The research workspaces above are built on one of the following templates. Each is shaped around a different research mode — see Choosing a template below.
Claude Research Space Public Template
Heavyweight workflow for iterative, multi-round AI-assisted research projects intended to be published. Includes a prompt queue, individual/aggregated outputs, compaction-based context management, voice-note ingestion (AssemblyAI), and built-in /export commands for blog posts, reports, briefings, and social threads. Use when the deliverable is a published research artefact and the process itself is the audit trail.
Claude Technical Research Template
Lightweight Q&A workspace for capturing technical how-to questions and their AI-assisted answers as a maintainable, multi-file reference — like a GitHub Gist, but versioned and built to be revised over time. Each topic is a cross-referenced pair: questions/<slug>.md (the question as posed) and ideas/<slug>.md (the long-form response). Includes /ask, /revise, /followup, /glossary, and /consolidate (Typst+PDF bundling). Use when the deliverable is a growing personal/public reference of "how do I do X?" answers, not a single research report.
Ecosystem Mapper
Claude Code plugin for delineating an emerging tech ecosystem — discovers, categorises, and publishes a structured GitHub repository mapping the communities, orgs, projects, and resources around a keyword (e.g. "agentic AI", "vector databases"). Includes skills for creating, updating, organising, and visualising the resulting map. Use when the deliverable is a curated landscape map rather than narrative research or Q&A.
Choosing a template
| If you're doing… | Use | Output shape |
|---|---|---|
| A single research project with rounds of prompts, compaction, and a polished published artefact | Claude Research Space Public Template | One report (with exports to blog/PDF/social), built up over multiple research rounds |
| Open-ended capture of technical Q&A as a living reference, one topic at a time | Claude Technical Research Template | A growing collection of question/idea pairs, optionally consolidated into PDFs |
| Mapping the players, projects, and resources in a tech ecosystem | Ecosystem Mapper | A categorised, badge-rich GitHub repo that functions as a landscape map |
Rough rule of thumb: Public Template is for one big question, many rounds, one publishable answer. Technical Research Template is for many small questions, one round each, one growing reference. Ecosystem Mapper is for one keyword, structured taxonomy, one curated map.
Related Indices
Research Tools Index
Claude Code research workspaces, templates, agents, and related tooling for deep-research workflows.
License
