A curated guide for LLM-agent-driven scientific research automation — from getting started to the frontier.
JavaScript
36
27 commits
updated Apr 21, 2026
README
Vibe Research Guide
A curated guide for LLM-agent-driven scientific research automation
🌐 View Interactive Multilingual README →
🇨🇳 中文 · 🇺🇸 English · 🇰🇷 한국어 · 🇯🇵 日本語 · 🇩🇪 Deutsch · 🇫🇷 Français · 🇪🇸 Español · 🇮🇹 Italiano · 🇵🇹 Português · 🇸🇦 العربية · 🇹🇭 ไทย · 🇻🇳 Tiếng Việt · 🇷🇺 Русский
Vibe Research At The Center, With The Broader Agent-Native Stack Around It
Automate the research loop with LLM agents: literature review → idea generation → experiment execution → paper writing → peer review.
This repo is a research-first landing page for the field. The center is still Vibe Research, but the guide now gives Auto Research / AI Scientist its own top-level map and then places Claw, coding agents, connectors, and adjacent assistant ecosystems around that core.
Start here: Getting Started · Auto Research · Tools & Platforms · Claw Park
At A Glance
|
Core Question How far can AI move from research assistant to research operator? Focus: literature, ideation, experiment, writing, and evaluation. |
What Changed In 2026 Research copilots got stronger, learning layers became real, autonomous research systems got more credible, and Vibe Coding became the execution layer. |
How To Use This Repo Treat the README as a map. Treat the topic pages as the actual guide. |
2026 Landscape Snapshot
Five shifts now define the field:
- Research copilots are stronger and easier to trust: Deep Research, NotebookLM-style source-grounded reading, and scientific workspaces such as Prism are making synthesis and report-writing meaningfully faster.
- Auto Research is becoming a recognizable system category: The AI Scientist, The AI Scientist-v2, Agent Laboratory, AI-Researcher, RD-Agent, Auto-Deep-Research, and EvoScientist now form a real system family.
- Benchmarks are getting closer to scientific reality: ScienceAgentBench, FIRE-Bench, ResearchClawBench, SGI-Bench, and RE-Bench shift evaluation away from vague "agent capability" claims.
- Execution substrates now determine whether research agents actually work: Claude Code, Codex, OpenHands, SWE-agent, OpenClaw, and chat bridges such as cc-connect increasingly matter because experiment loops fail on tooling long before they fail on prose.
- A wider agent-native ecosystem is forming around research: personal assistants, registries, skills, self-evolving stacks, and companion UX are increasingly part of the same operational environment.
2026 Auto Research Signals
Several current signals make the field feel less like a loose collection of demos and more like an emerging research stack:
- The field has moved from "can an agent summarize papers?" to "can an agent operate a research loop?"
- System families are diverging clearly: end-to-end scientists, human-in-the-loop research copilots, R&D execution agents, deep-research assistants, and self-evolving scientist stacks now look meaningfully different.
- Benchmarking is improving fast: scientist-aligned workflows, rediscovery tasks, expert comparison, and checklist-based evaluation are becoming standard expectations instead of afterthoughts.
- Frameworks matter as much as flagship demos: AI-Researcher, RD-Agent, and Auto-Deep-Research show that orchestration and harness design are now first-class concerns.
- Platformization is visible: FutureHouse Platform, Robin, BixBench, and Edison Scientific / Kosmos show how AI-scientist ideas are moving from one-off papers to persistent public or commercial surfaces.
Choose a Path
|
🟢 New to Vibe Research Start: Getting Started Then: Auto Research · Tools & Platforms |
🔵 Developer / Builder Start: Auto Research Then: Tools & Platforms · Vibe Coding · Systems |
|
🔴 Researcher Start: Surveys Then: Auto Research · Benchmarks · Ideation |
🟣 Creator / Operator Start: Tools & Platforms Then: Vibe Coding · Vibe Anything |
Only have 5 minutes? Install InnoClaw and try it out.
Auto Research Stack
This is the shortest useful way to read the field in 2026:
| Layer | Representative resources | Why it matters |
|---|---|---|
| Research copilots | Deep Research · NotebookLM · Prism | Best entry point for synthesis, reading, and report generation |
| Auto Research systems | The AI Scientist · The AI Scientist-v2 · Agent Laboratory · EvoScientist | Defines what end-to-end or near-end-to-end research automation looks like |
| Orchestration frameworks | AI-Researcher · RD-Agent · Auto-Deep-Research | Shows the framework layer growing around research loops, not only one-off papers |
| Benchmarks & scientist-aligned eval | ScienceAgentBench · FIRE-Bench · ResearchClawBench · SGI-Bench · RE-Bench | Keeps the field grounded in rediscovery, expert comparison, and workflow realism |
| Execution substrate | Claude Code · Codex · OpenHands · SWE-agent · OpenClaw | Most research-agent failures now happen here: tool use, code execution, environment control, and iteration |
| Platform signals | FutureHouse Platform · Robin · Edison Scientific | Shows the move from repos and papers toward durable product surfaces |
For a dedicated overview, start here: → Auto Research
Representative Auto Research Systems
| Family | Representative resources | What it optimizes for |
|---|---|---|
| End-to-end AI scientist | The AI Scientist · The AI Scientist-v2 | Idea generation, experiment execution, and paper/report production |
| Human-in-the-loop research copilot | Agent Laboratory | Collaboration and controllable research assistance |
| Research orchestration framework | AI-Researcher · Auto-Deep-Research | Search, synthesis, and multi-stage research workflow control |
| Autonomous R&D / data science | RD-Agent | Real implementation loops, evaluation, and applied experimentation |
| Self-evolving scientist | EvoScientist | Memory, iteration, and continual improvement of the research process itself |
Auto Research Benchmarks
| Benchmark | What it measures | Why it matters |
|---|---|---|
| ScienceAgentBench | Scientific discovery tasks with grounded evaluation | One of the clearest early attempts to benchmark real research-agent capability |
| FIRE-Bench | Rediscovery of known scientific insights | Makes "can the system rediscover something real?" a first-class metric |
| ResearchClawBench | Autonomous research from rediscovery to new-discovery | Strong recent signal that agentic research-workspace evaluation is becoming more realistic |
| SGI-Bench | Scientist-aligned workflows and scientific general intelligence | Separates scientist-like process quality from generic language-model fluency |
| RE-Bench | Frontier AI R&D against human experts | Useful reality check for how far autonomous R&D actually is from human performance |
More detail: → Benchmarks
Agent-Native Landscape Beyond Research
Vibe Research remains the center of this repo. But in practice, research agents now live inside a larger operating environment: personal assistants, software-surface layers, self-improving agents, and companion UX around coding agents.
Personal Agent Assistants
These are not all "research agents", but they increasingly shape the environment research agents live in.
| Project | What it is | Why it matters here |
|---|---|---|
| OpenClaw | Gateway-native assistant runtime with chat control, plugins, bundles, and deployment surfaces | Shows how personal assistants, plugins, and research flows can share one substrate instead of staying as separate demos |
| Hermes Agent | General-purpose personal agent stack with gateway, CLI, plugins, skills, and long-running plans | Important signal that "personal assistant" is becoming a real open-source platform layer, not only a chatbot wrapper |
| Goose | Open-source extensible agent that can install, execute, edit, and test with any LLM | Represents the dev-native branch of personal assistants that sits close to repo work and engineering execution |
| Khoj | Self-hostable AI second brain and autonomous personal AI with deep-research and automation hooks | Shows the knowledge-native assistant pattern where long-term memory, personal docs, and web retrieval become part of the assistant surface |
| AnythingLLM | Privacy-first workspace-style AI productivity layer | Useful reference for the workspace-native assistant pattern where teams want one local-first surface for models, docs, and agents |
Agent-Native Software / CLI-Anything
This layer matters because the frontier is no longer only "which agent is best", but also "which software surfaces are now agent-operable".
| Project | What it is | Why it matters here |
|---|---|---|
| CLI-Anything | Turns software into agent-native CLI surfaces and ships many agent-harness adapters | Strong signal that existing tools are being retrofitted into agent-operable interfaces instead of being rebuilt from scratch |
| cc-connect | Chat control plane for Claude Code, Codex, Cursor, Gemini CLI, and related agents | Shows how chat surfaces are becoming remote-control shells for coding and research agents |
| Official MCP Registry | Canonical discovery layer for MCP servers and tools | Makes tool discovery and installation a real infrastructure concern instead of ad hoc glue |
| anthropics/skills | Public reusable skill substrate for Claude Code | Shows how skills are becoming portable capability units rather than only local prompts |
| ClawHub · OpenClaw Plugin Bundles | Registry, bundle compatibility, and install layer around OpenClaw | Good reference for how agent-native ecosystems package and distribute capabilities |
Learning, RL & Self-Evolving Agents
This is still one of the most important structural shifts in the field: not just tool-using agents, but agents that can be trained, optimized, or improved over time.
| Layer | Representative resources | Why it matters |
|---|---|---|
| Agent training / optimization | Agent Lightning | Brings RL, automatic prompt optimization, and SFT to arbitrary agent systems with near-zero code changes |
| Zero-data self-evolution | Agent0 · AgentEvolver | Shows how agents can generate tasks, feedback, and training signals without human-curated data pipelines |
| Evolving workflows | EvoAgentX · EvoScientist · MetaClaw | Shifts the focus from optimizing one prompt to evolving whole workflows, skill graphs, or scientist loops |
| Skill & memory substrate | Acontext · anthropics/skills | Makes skills, context, and reusable experience part of the learning layer |
| Landscape map | Awesome-Self-Evolving-Agents | Best current GitHub-native overview of the optimization, evolution, and lifelong-agent literature |
Vibe Coding Apps & Companion UX
This is a newer branch, but it is growing quickly enough that it deserves to be called out separately.
| Project | What it is | Why it matters here |
|---|---|---|
| Crush | Open-source agentic coding UX from Charm | Good signal that the UI layer around coding agents is now its own product surface, not just a wrapper around a model API |
| Vibe Island | Commercial macOS Dynamic Island companion for many coding agents | Shows the new monitor / approve / ask / jump-back interaction layer around long-running coding-agent sessions |
| xisland | Free macOS notch companion for Claude Code, Codex, Gemini CLI, and OpenCode | Similar signal from an indie distribution angle: approval and session-monitoring UX is becoming a category of its own |
Claw Stack At A Glance
The Claw family now spans more than research agents. The cleaner reading is:
| Layer | Representative projects | Why it matters |
|---|---|---|
| Gateway / foundation | OpenClaw | Core runtime, control surface, and the base layer other Claw projects increasingly build on |
| Registry / discovery | ClawHub · awesome-openclaw-skills | Skill and plugin discovery are now a separate ecosystem layer |
| Compatibility / bundles | OpenClaw Plugin Bundles | Makes it easier to import Codex, Claude, and Cursor ecosystem formats into OpenClaw-native features |
| Packaging / deployment | nix-openclaw | Gives the ecosystem a reproducible deployment path for serious self-hosting and ops |
| Research surfaces | InnoClaw · ResearchClaw · ResearchClaw Desktop App | Covers grounded workspaces, daily research copilots, and lighter-weight reading surfaces |
| Scientific specialist / evolution | ScienceClaw · MetaClaw | Pushes deeper into scientific specialization, persistent memory, and online learning |
| Autonomous pipeline | AutoResearchClaw | Represents the maximum-autonomy idea-to-paper direction |
Full map: → Claw Park
Ecosystem Snapshot
| Layer | Representative projects | Why it matters |
|---|---|---|
| Research copilots | OpenAI Deep Research · Gemini Deep Research · NotebookLM · Prism | Fast literature synthesis, source-grounded reading, and scientific writing assistance |
| Auto Research systems | The AI Scientist · The AI Scientist-v2 · Agent Laboratory · AI-Researcher · RD-Agent · Auto-Deep-Research · EvoScientist | The current system family for automating or orchestrating meaningful parts of the research loop |
| Scientist-aligned benchmarks | ScienceAgentBench · FIRE-Bench · ResearchClawBench · SGI-Bench · RE-Bench | Keeps the field grounded in rediscovery, expert comparison, and research-workflow realism |
| AI scientist platforms | FutureHouse Platform · Robin · Edison Scientific · Kosmos | Shows the field moving from paper demos to persistent web/API platforms and validated scientific workflows |
| Research workspaces | InnoClaw · ResearchClaw · FARS · OpenClaw | Workspaces and assistant shells that keep papers, files, notes, code, and local execution in one loop |
| Learning / self-evolving layer | Agent Lightning · Agent0 · AgentEvolver · EvoAgentX · Acontext | Turns agent training, self-generated data, evolving workflows, and persistent skill/context memory into a real stack layer |
| Agent-native software / harnesses | CLI-Anything · cc-connect · Official MCP Registry · anthropics/skills | Shows how software surfaces, registries, and skills are becoming agent-operable infrastructure |
| Claw ecosystem | OpenClaw · ClawHub · OpenClaw Plugin Bundles · nix-openclaw · InnoClaw · ResearchClaw · ScienceClaw · MetaClaw · AutoResearchClaw | Gateway, registry, compatibility, deployment, research workspaces, scientific specialization, online learning, and autonomous pipelines |
| Execution layer | Claude Code · Codex · Cursor Background Agents · GitHub Copilot Coding Agent | The coding and repo workflow layer that increasingly powers research execution |
| Personal agent assistants | Hermes Agent · Goose · Khoj · AnythingLLM | Shows the wider assistant layer surrounding research: messaging-native, dev-native, knowledge-native, and workspace-native agents |
| Companion apps / coding UX | Crush · Vibe Island · xisland | Shows the monitoring, approval, and session-jump UX layer forming around long-running coding agents |
| Adjacent prompt-native tools | v0 · Lovable · Replit Agent | Useful for prototyping, but not the core of Vibe Research |
Plugins, Bridges & Research Connectors
A new layer is forming between "agent" and "workflow": plugin surfaces, MCP registries, skill catalogs, and chat bridges that make research agents easier to extend, discover, and operate.
| Layer | Representative resources | Why it matters |
|---|---|---|
| Bridge & control surfaces | cc-connect | Runs Claude Code, Cursor, Gemini CLI, Codex, and similar agents from chat surfaces such as Feishu/Lark, Slack, Telegram, and WeCom |
| Plugin / customization layer | CLI-Anything · ClawHub · OpenClaw Plugin Bundles · awesome-claude-code-plugins | Shows how agent ecosystems are moving toward software harnesses, skill registries, plugin marketplaces, bundle compatibility, and installable capability packs |
| Learning / memory substrate | Acontext · anthropics/skills | Shows how context, memory, and reusable skills are turning into persistent substrates for agent improvement |
| Claude Code workflow layer | wshobson/agents · SuperClaude Framework · claude-task-master | Shows how commands, agent teams, skills, and task systems are turning Claude Code into a fuller development environment |
| Routing / agent-ops layer | claude-code-router · Claude Squad · Repomix | Highlights provider routing, multi-agent session management, and codebase packaging as new operational layers around coding agents |
| Registry / discovery layer | Official MCP Registry · awesome-mcp-servers · awesome-openclaw-skills | Makes it easier to find, compare, and install the rapidly growing tool and skill ecosystem |
| Research connectors | OpenAlex Research MCP · Academia MCP · PapersWithCode MCP | Connects agents directly to literature graphs, code artifacts, datasets, and benchmark metadata |
More detailed map: → Tools & Platforms
Topic Map
Core Guides
| Topic | Description | Link |
|---|---|---|
| 🚀 Getting Started | 5-min demo → 30-min agent deployment → full automation | → Getting Started |
| 🔬 Auto Research | Map the AI Scientist stack across papers, systems, benchmarks, frameworks, and platform signals | → Auto Research |
| 🧰 Tools & Platforms | Core research platforms plus assistant, connector, and software-surface layers | → Tools & Platforms |
| 🦞 Claw Park | Ecosystem map for what each Claw project is building and where it fits | → Claw Park |
| 💻 Vibe Coding | Terminal agents, coding agents, companion apps, and repo guardrails | → Vibe Coding |
| 🎨 Vibe Anything | Adjacent prompt-native workflows for apps, design, writing, slides, and ops | → Vibe Anything |
Research Topics (39+ papers)
| Topic | Core Question | Papers | Link |
|---|---|---|---|
| 📄 Surveys | Landscape, scientific-method framing, and evolution of the field | 6 | → Surveys |
| ⚙️ Systems | How to design end-to-end research systems | 7 | → Systems |
| 💡 Ideation | Can LLMs generate novel ideas | 6 | → Ideation |
| 📚 Synthesis | How to synthesize literature at scale | 5 | → Synthesis |
| 🧪 Experiment | How agents automate experiments | 4 | → Experiment |
| ✍️ Writing & Review | LLM-assisted writing & peer review | 4 | → Writing & Review |
| 📊 Benchmarks | How to evaluate research agents | 7 | → Benchmarks |
Reading Modes
|
Read The Field Auto Research · Surveys · Systems · Benchmarks |
Build The Stack Tools & Platforms · Claw Park · Vibe Coding |
Prototype Beyond Research Vibe Anything |
Useful Resources
Introductions: AI for Science (Nature) · LLM Agents (Lilian Weng) · Agentic Patterns (Andrew Ng)
Awesome Lists: LLM Agent Survey · AI Agents · Scientific Idea Generation
Research Core: Semantic Scholar · Elicit · Consensus · Connected Papers
Auto Research / AI Scientist: The AI Scientist · The AI Scientist-v2 · Agent Laboratory · AI-Researcher · RD-Agent · Auto-Deep-Research · EvoScientist
Benchmarks & Evaluation: ScienceAgentBench · FIRE-Bench · ResearchClawBench · SGI-Bench · RE-Bench · MLE-bench
Scientific-Method Framing: From Automation to Autonomy · LLM Agents as AI Scientists: A Survey · A Survey of LLM-based Scientific Agents · Exploring the role of large language models in the scientific method
AI Scientist Platforms: FutureHouse Platform · Robin · BixBench · Edison Scientific · Kosmos
Personal Agent Assistants: OpenClaw · Hermes Agent · Goose · Khoj · AnythingLLM
Agent-Native Interfaces & Harnesses: CLI-Anything · cc-connect · Official MCP Registry · anthropics/skills · ClawHub
Claw Ecosystem: OpenClaw · ClawHub · Plugin Bundles · nix-openclaw · Claw Park
Learning / Self-Evolving Agents: Agent Lightning · Agent0 · AgentEvolver · EvoAgentX · Acontext · Awesome-Self-Evolving-Agents
Execution / Vibe Coding: Claude Code · Codex · Cursor Background Agents · GitHub Copilot Coding Agent · Gemini CLI · Crush
Claude Code Ecosystem: anthropics/skills · wshobson/agents · SuperClaude Framework · claude-code-router · Claude Squad · claude-task-master · Repomix
Companion Apps: Vibe Island · xisland
Prototyping: v0 · Lovable · Replit Agent · Figma AI · Canva AI
Conferences: NeurIPS · ICML · ICLR · ACL · AAAI · EMNLP
Contribute
Submit resources via Resource Suggestion · Contribute via PR · Follow the curation guidelines
Citation
@misc{viberesearch2026,
title = {Vibe Research Guide},
author = {Aaron Wang and Contributors},
year = {2026},
url = {https://github.com/SpectrAI-Initiative/Vibe-Research-Guide},
}
Changelog
- 2026-04-21: Added a dedicated
Auto Researchcore guide, reweighted the README back towardVibe Research, refreshedSystems / Surveys / Benchmarks / Tools, and synced the landing page to the stronger auto-research framing - 2026-04-15: Reframed the README as a research-first but broader agent-native map, adding personal assistants, CLI-Anything / harness layers, and companion coding apps while keeping
Vibe Researchas the center - 2026-W17: Expanded Claw coverage from a short project list into a fuller family map, including ClawHub, Plugin Bundles, nix-openclaw, ResearchClaw Desktop App, and a clearer stack-layer taxonomy
- 2026-W16: Added a dedicated learning / RL / self-evolving layer to the guide, including Agent Lightning, Agent0, AgentEvolver, EvoAgentX, Acontext, and Awesome-Self-Evolving-Agents
- 2026-W14: Added 2026 Q1 signals for OpenClaw platformization, FutureHouse / Robin / BixBench, and Edison Scientific / Kosmos; refreshed ecosystem framing across the guide
- 2026-W14: Added recent Claude Code ecosystem signals, including anthropics/skills, wshobson/agents, SuperClaude, claude-code-router, Claude Squad, claude-task-master, and Repomix
- 2026-W13: Added a new plugin / bridge / registry layer to the guide, including cc-connect, OpenAlex Research MCP, Academia MCP, PapersWithCode MCP, and more Claw ecosystem positioning
- 2026-W13: Added core tools & platforms (InnoClaw, ResearchClaw, FARS, Orchestra, OpenClaw, EvoScientist); added Deep Research tools, OpenAI Prism, MCP Servers; switched all content to English; expanded to 35+ papers across 9 topic files
- 2026-W12: Redesigned README into a stronger landing page with cleaner hierarchy, card-style path selection, and a more visual ecosystem map
- 2026-W12: Hub-and-spoke architecture reorganization
- 2026-W12: Initial public release
Full history: CHANGELOG.md
MIT License ·
ai-science
literature-review
llm-agents
scientific-discovery
vibe-research
Significant stargazers
Seungwoo hong
98 followers · starred May 2026
Similar repos
A curated guide for LLM-agent-driven scientific research automation — from getting started to the frontier.
JavaScript
36
27 commits
updated Apr 21, 2026
README
Vibe Research Guide
A curated guide for LLM-agent-driven scientific research automation
🌐 View Interactive Multilingual README →
🇨🇳 中文 · 🇺🇸 English · 🇰🇷 한국어 · 🇯🇵 日本語 · 🇩🇪 Deutsch · 🇫🇷 Français · 🇪🇸 Español · 🇮🇹 Italiano · 🇵🇹 Português · 🇸🇦 العربية · 🇹🇭 ไทย · 🇻🇳 Tiếng Việt · 🇷🇺 Русский
Vibe Research At The Center, With The Broader Agent-Native Stack Around It
Automate the research loop with LLM agents: literature review → idea generation → experiment execution → paper writing → peer review.
This repo is a research-first landing page for the field. The center is still Vibe Research, but the guide now gives Auto Research / AI Scientist its own top-level map and then places Claw, coding agents, connectors, and adjacent assistant ecosystems around that core.
Start here: Getting Started · Auto Research · Tools & Platforms · Claw Park
At A Glance
|
Core Question How far can AI move from research assistant to research operator? Focus: literature, ideation, experiment, writing, and evaluation. |
What Changed In 2026 Research copilots got stronger, learning layers became real, autonomous research systems got more credible, and Vibe Coding became the execution layer. |
How To Use This Repo Treat the README as a map. Treat the topic pages as the actual guide. |
2026 Landscape Snapshot
Five shifts now define the field:
- Research copilots are stronger and easier to trust: Deep Research, NotebookLM-style source-grounded reading, and scientific workspaces such as Prism are making synthesis and report-writing meaningfully faster.
- Auto Research is becoming a recognizable system category: The AI Scientist, The AI Scientist-v2, Agent Laboratory, AI-Researcher, RD-Agent, Auto-Deep-Research, and EvoScientist now form a real system family.
- Benchmarks are getting closer to scientific reality: ScienceAgentBench, FIRE-Bench, ResearchClawBench, SGI-Bench, and RE-Bench shift evaluation away from vague "agent capability" claims.
- Execution substrates now determine whether research agents actually work: Claude Code, Codex, OpenHands, SWE-agent, OpenClaw, and chat bridges such as cc-connect increasingly matter because experiment loops fail on tooling long before they fail on prose.
- A wider agent-native ecosystem is forming around research: personal assistants, registries, skills, self-evolving stacks, and companion UX are increasingly part of the same operational environment.
2026 Auto Research Signals
Several current signals make the field feel less like a loose collection of demos and more like an emerging research stack:
- The field has moved from "can an agent summarize papers?" to "can an agent operate a research loop?"
- System families are diverging clearly: end-to-end scientists, human-in-the-loop research copilots, R&D execution agents, deep-research assistants, and self-evolving scientist stacks now look meaningfully different.
- Benchmarking is improving fast: scientist-aligned workflows, rediscovery tasks, expert comparison, and checklist-based evaluation are becoming standard expectations instead of afterthoughts.
- Frameworks matter as much as flagship demos: AI-Researcher, RD-Agent, and Auto-Deep-Research show that orchestration and harness design are now first-class concerns.
- Platformization is visible: FutureHouse Platform, Robin, BixBench, and Edison Scientific / Kosmos show how AI-scientist ideas are moving from one-off papers to persistent public or commercial surfaces.
Choose a Path
|
🟢 New to Vibe Research Start: Getting Started Then: Auto Research · Tools & Platforms |
🔵 Developer / Builder Start: Auto Research Then: Tools & Platforms · Vibe Coding · Systems |
|
🔴 Researcher Start: Surveys Then: Auto Research · Benchmarks · Ideation |
🟣 Creator / Operator Start: Tools & Platforms Then: Vibe Coding · Vibe Anything |
Only have 5 minutes? Install InnoClaw and try it out.
Auto Research Stack
This is the shortest useful way to read the field in 2026:
| Layer | Representative resources | Why it matters |
|---|---|---|
| Research copilots | Deep Research · NotebookLM · Prism | Best entry point for synthesis, reading, and report generation |
| Auto Research systems | The AI Scientist · The AI Scientist-v2 · Agent Laboratory · EvoScientist | Defines what end-to-end or near-end-to-end research automation looks like |
| Orchestration frameworks | AI-Researcher · RD-Agent · Auto-Deep-Research | Shows the framework layer growing around research loops, not only one-off papers |
| Benchmarks & scientist-aligned eval | ScienceAgentBench · FIRE-Bench · ResearchClawBench · SGI-Bench · RE-Bench | Keeps the field grounded in rediscovery, expert comparison, and workflow realism |
| Execution substrate | Claude Code · Codex · OpenHands · SWE-agent · OpenClaw | Most research-agent failures now happen here: tool use, code execution, environment control, and iteration |
| Platform signals | FutureHouse Platform · Robin · Edison Scientific | Shows the move from repos and papers toward durable product surfaces |
For a dedicated overview, start here: → Auto Research
Representative Auto Research Systems
| Family | Representative resources | What it optimizes for |
|---|---|---|
| End-to-end AI scientist | The AI Scientist · The AI Scientist-v2 | Idea generation, experiment execution, and paper/report production |
| Human-in-the-loop research copilot | Agent Laboratory | Collaboration and controllable research assistance |
| Research orchestration framework | AI-Researcher · Auto-Deep-Research | Search, synthesis, and multi-stage research workflow control |
| Autonomous R&D / data science | RD-Agent | Real implementation loops, evaluation, and applied experimentation |
| Self-evolving scientist | EvoScientist | Memory, iteration, and continual improvement of the research process itself |
Auto Research Benchmarks
| Benchmark | What it measures | Why it matters |
|---|---|---|
| ScienceAgentBench | Scientific discovery tasks with grounded evaluation | One of the clearest early attempts to benchmark real research-agent capability |
| FIRE-Bench | Rediscovery of known scientific insights | Makes "can the system rediscover something real?" a first-class metric |
| ResearchClawBench | Autonomous research from rediscovery to new-discovery | Strong recent signal that agentic research-workspace evaluation is becoming more realistic |
| SGI-Bench | Scientist-aligned workflows and scientific general intelligence | Separates scientist-like process quality from generic language-model fluency |
| RE-Bench | Frontier AI R&D against human experts | Useful reality check for how far autonomous R&D actually is from human performance |
More detail: → Benchmarks
Agent-Native Landscape Beyond Research
Vibe Research remains the center of this repo. But in practice, research agents now live inside a larger operating environment: personal assistants, software-surface layers, self-improving agents, and companion UX around coding agents.
Personal Agent Assistants
These are not all "research agents", but they increasingly shape the environment research agents live in.
| Project | What it is | Why it matters here |
|---|---|---|
| OpenClaw | Gateway-native assistant runtime with chat control, plugins, bundles, and deployment surfaces | Shows how personal assistants, plugins, and research flows can share one substrate instead of staying as separate demos |
| Hermes Agent | General-purpose personal agent stack with gateway, CLI, plugins, skills, and long-running plans | Important signal that "personal assistant" is becoming a real open-source platform layer, not only a chatbot wrapper |
| Goose | Open-source extensible agent that can install, execute, edit, and test with any LLM | Represents the dev-native branch of personal assistants that sits close to repo work and engineering execution |
| Khoj | Self-hostable AI second brain and autonomous personal AI with deep-research and automation hooks | Shows the knowledge-native assistant pattern where long-term memory, personal docs, and web retrieval become part of the assistant surface |
| AnythingLLM | Privacy-first workspace-style AI productivity layer | Useful reference for the workspace-native assistant pattern where teams want one local-first surface for models, docs, and agents |
Agent-Native Software / CLI-Anything
This layer matters because the frontier is no longer only "which agent is best", but also "which software surfaces are now agent-operable".
| Project | What it is | Why it matters here |
|---|---|---|
| CLI-Anything | Turns software into agent-native CLI surfaces and ships many agent-harness adapters | Strong signal that existing tools are being retrofitted into agent-operable interfaces instead of being rebuilt from scratch |
| cc-connect | Chat control plane for Claude Code, Codex, Cursor, Gemini CLI, and related agents | Shows how chat surfaces are becoming remote-control shells for coding and research agents |
| Official MCP Registry | Canonical discovery layer for MCP servers and tools | Makes tool discovery and installation a real infrastructure concern instead of ad hoc glue |
| anthropics/skills | Public reusable skill substrate for Claude Code | Shows how skills are becoming portable capability units rather than only local prompts |
| ClawHub · OpenClaw Plugin Bundles | Registry, bundle compatibility, and install layer around OpenClaw | Good reference for how agent-native ecosystems package and distribute capabilities |
Learning, RL & Self-Evolving Agents
This is still one of the most important structural shifts in the field: not just tool-using agents, but agents that can be trained, optimized, or improved over time.
| Layer | Representative resources | Why it matters |
|---|---|---|
| Agent training / optimization | Agent Lightning | Brings RL, automatic prompt optimization, and SFT to arbitrary agent systems with near-zero code changes |
| Zero-data self-evolution | Agent0 · AgentEvolver | Shows how agents can generate tasks, feedback, and training signals without human-curated data pipelines |
| Evolving workflows | EvoAgentX · EvoScientist · MetaClaw | Shifts the focus from optimizing one prompt to evolving whole workflows, skill graphs, or scientist loops |
| Skill & memory substrate | Acontext · anthropics/skills | Makes skills, context, and reusable experience part of the learning layer |
| Landscape map | Awesome-Self-Evolving-Agents | Best current GitHub-native overview of the optimization, evolution, and lifelong-agent literature |
Vibe Coding Apps & Companion UX
This is a newer branch, but it is growing quickly enough that it deserves to be called out separately.
| Project | What it is | Why it matters here |
|---|---|---|
| Crush | Open-source agentic coding UX from Charm | Good signal that the UI layer around coding agents is now its own product surface, not just a wrapper around a model API |
| Vibe Island | Commercial macOS Dynamic Island companion for many coding agents | Shows the new monitor / approve / ask / jump-back interaction layer around long-running coding-agent sessions |
| xisland | Free macOS notch companion for Claude Code, Codex, Gemini CLI, and OpenCode | Similar signal from an indie distribution angle: approval and session-monitoring UX is becoming a category of its own |
Claw Stack At A Glance
The Claw family now spans more than research agents. The cleaner reading is:
| Layer | Representative projects | Why it matters |
|---|---|---|
| Gateway / foundation | OpenClaw | Core runtime, control surface, and the base layer other Claw projects increasingly build on |
| Registry / discovery | ClawHub · awesome-openclaw-skills | Skill and plugin discovery are now a separate ecosystem layer |
| Compatibility / bundles | OpenClaw Plugin Bundles | Makes it easier to import Codex, Claude, and Cursor ecosystem formats into OpenClaw-native features |
| Packaging / deployment | nix-openclaw | Gives the ecosystem a reproducible deployment path for serious self-hosting and ops |
| Research surfaces | InnoClaw · ResearchClaw · ResearchClaw Desktop App | Covers grounded workspaces, daily research copilots, and lighter-weight reading surfaces |
| Scientific specialist / evolution | ScienceClaw · MetaClaw | Pushes deeper into scientific specialization, persistent memory, and online learning |
| Autonomous pipeline | AutoResearchClaw | Represents the maximum-autonomy idea-to-paper direction |
Full map: → Claw Park
Ecosystem Snapshot
| Layer | Representative projects | Why it matters |
|---|---|---|
| Research copilots | OpenAI Deep Research · Gemini Deep Research · NotebookLM · Prism | Fast literature synthesis, source-grounded reading, and scientific writing assistance |
| Auto Research systems | The AI Scientist · The AI Scientist-v2 · Agent Laboratory · AI-Researcher · RD-Agent · Auto-Deep-Research · EvoScientist | The current system family for automating or orchestrating meaningful parts of the research loop |
| Scientist-aligned benchmarks | ScienceAgentBench · FIRE-Bench · ResearchClawBench · SGI-Bench · RE-Bench | Keeps the field grounded in rediscovery, expert comparison, and research-workflow realism |
| AI scientist platforms | FutureHouse Platform · Robin · Edison Scientific · Kosmos | Shows the field moving from paper demos to persistent web/API platforms and validated scientific workflows |
| Research workspaces | InnoClaw · ResearchClaw · FARS · OpenClaw | Workspaces and assistant shells that keep papers, files, notes, code, and local execution in one loop |
| Learning / self-evolving layer | Agent Lightning · Agent0 · AgentEvolver · EvoAgentX · Acontext | Turns agent training, self-generated data, evolving workflows, and persistent skill/context memory into a real stack layer |
| Agent-native software / harnesses | CLI-Anything · cc-connect · Official MCP Registry · anthropics/skills | Shows how software surfaces, registries, and skills are becoming agent-operable infrastructure |
| Claw ecosystem | OpenClaw · ClawHub · OpenClaw Plugin Bundles · nix-openclaw · InnoClaw · ResearchClaw · ScienceClaw · MetaClaw · AutoResearchClaw | Gateway, registry, compatibility, deployment, research workspaces, scientific specialization, online learning, and autonomous pipelines |
| Execution layer | Claude Code · Codex · Cursor Background Agents · GitHub Copilot Coding Agent | The coding and repo workflow layer that increasingly powers research execution |
| Personal agent assistants | Hermes Agent · Goose · Khoj · AnythingLLM | Shows the wider assistant layer surrounding research: messaging-native, dev-native, knowledge-native, and workspace-native agents |
| Companion apps / coding UX | Crush · Vibe Island · xisland | Shows the monitoring, approval, and session-jump UX layer forming around long-running coding agents |
| Adjacent prompt-native tools | v0 · Lovable · Replit Agent | Useful for prototyping, but not the core of Vibe Research |
Plugins, Bridges & Research Connectors
A new layer is forming between "agent" and "workflow": plugin surfaces, MCP registries, skill catalogs, and chat bridges that make research agents easier to extend, discover, and operate.
| Layer | Representative resources | Why it matters |
|---|---|---|
| Bridge & control surfaces | cc-connect | Runs Claude Code, Cursor, Gemini CLI, Codex, and similar agents from chat surfaces such as Feishu/Lark, Slack, Telegram, and WeCom |
| Plugin / customization layer | CLI-Anything · ClawHub · OpenClaw Plugin Bundles · awesome-claude-code-plugins | Shows how agent ecosystems are moving toward software harnesses, skill registries, plugin marketplaces, bundle compatibility, and installable capability packs |
| Learning / memory substrate | Acontext · anthropics/skills | Shows how context, memory, and reusable skills are turning into persistent substrates for agent improvement |
| Claude Code workflow layer | wshobson/agents · SuperClaude Framework · claude-task-master | Shows how commands, agent teams, skills, and task systems are turning Claude Code into a fuller development environment |
| Routing / agent-ops layer | claude-code-router · Claude Squad · Repomix | Highlights provider routing, multi-agent session management, and codebase packaging as new operational layers around coding agents |
| Registry / discovery layer | Official MCP Registry · awesome-mcp-servers · awesome-openclaw-skills | Makes it easier to find, compare, and install the rapidly growing tool and skill ecosystem |
| Research connectors | OpenAlex Research MCP · Academia MCP · PapersWithCode MCP | Connects agents directly to literature graphs, code artifacts, datasets, and benchmark metadata |
More detailed map: → Tools & Platforms
Topic Map
Core Guides
| Topic | Description | Link |
|---|---|---|
| 🚀 Getting Started | 5-min demo → 30-min agent deployment → full automation | → Getting Started |
| 🔬 Auto Research | Map the AI Scientist stack across papers, systems, benchmarks, frameworks, and platform signals | → Auto Research |
| 🧰 Tools & Platforms | Core research platforms plus assistant, connector, and software-surface layers | → Tools & Platforms |
| 🦞 Claw Park | Ecosystem map for what each Claw project is building and where it fits | → Claw Park |
| 💻 Vibe Coding | Terminal agents, coding agents, companion apps, and repo guardrails | → Vibe Coding |
| 🎨 Vibe Anything | Adjacent prompt-native workflows for apps, design, writing, slides, and ops | → Vibe Anything |
Research Topics (39+ papers)
| Topic | Core Question | Papers | Link |
|---|---|---|---|
| 📄 Surveys | Landscape, scientific-method framing, and evolution of the field | 6 | → Surveys |
| ⚙️ Systems | How to design end-to-end research systems | 7 | → Systems |
| 💡 Ideation | Can LLMs generate novel ideas | 6 | → Ideation |
| 📚 Synthesis | How to synthesize literature at scale | 5 | → Synthesis |
| 🧪 Experiment | How agents automate experiments | 4 | → Experiment |
| ✍️ Writing & Review | LLM-assisted writing & peer review | 4 | → Writing & Review |
| 📊 Benchmarks | How to evaluate research agents | 7 | → Benchmarks |
Reading Modes
|
Read The Field Auto Research · Surveys · Systems · Benchmarks |
Build The Stack Tools & Platforms · Claw Park · Vibe Coding |
Prototype Beyond Research Vibe Anything |
Useful Resources
Introductions: AI for Science (Nature) · LLM Agents (Lilian Weng) · Agentic Patterns (Andrew Ng)
Awesome Lists: LLM Agent Survey · AI Agents · Scientific Idea Generation
Research Core: Semantic Scholar · Elicit · Consensus · Connected Papers
Auto Research / AI Scientist: The AI Scientist · The AI Scientist-v2 · Agent Laboratory · AI-Researcher · RD-Agent · Auto-Deep-Research · EvoScientist
Benchmarks & Evaluation: ScienceAgentBench · FIRE-Bench · ResearchClawBench · SGI-Bench · RE-Bench · MLE-bench
Scientific-Method Framing: From Automation to Autonomy · LLM Agents as AI Scientists: A Survey · A Survey of LLM-based Scientific Agents · Exploring the role of large language models in the scientific method
AI Scientist Platforms: FutureHouse Platform · Robin · BixBench · Edison Scientific · Kosmos
Personal Agent Assistants: OpenClaw · Hermes Agent · Goose · Khoj · AnythingLLM
Agent-Native Interfaces & Harnesses: CLI-Anything · cc-connect · Official MCP Registry · anthropics/skills · ClawHub
Claw Ecosystem: OpenClaw · ClawHub · Plugin Bundles · nix-openclaw · Claw Park
Learning / Self-Evolving Agents: Agent Lightning · Agent0 · AgentEvolver · EvoAgentX · Acontext · Awesome-Self-Evolving-Agents
Execution / Vibe Coding: Claude Code · Codex · Cursor Background Agents · GitHub Copilot Coding Agent · Gemini CLI · Crush
Claude Code Ecosystem: anthropics/skills · wshobson/agents · SuperClaude Framework · claude-code-router · Claude Squad · claude-task-master · Repomix
Companion Apps: Vibe Island · xisland
Prototyping: v0 · Lovable · Replit Agent · Figma AI · Canva AI
Conferences: NeurIPS · ICML · ICLR · ACL · AAAI · EMNLP
Contribute
Submit resources via Resource Suggestion · Contribute via PR · Follow the curation guidelines
Citation
@misc{viberesearch2026,
title = {Vibe Research Guide},
author = {Aaron Wang and Contributors},
year = {2026},
url = {https://github.com/SpectrAI-Initiative/Vibe-Research-Guide},
}
Changelog
- 2026-04-21: Added a dedicated
Auto Researchcore guide, reweighted the README back towardVibe Research, refreshedSystems / Surveys / Benchmarks / Tools, and synced the landing page to the stronger auto-research framing - 2026-04-15: Reframed the README as a research-first but broader agent-native map, adding personal assistants, CLI-Anything / harness layers, and companion coding apps while keeping
Vibe Researchas the center - 2026-W17: Expanded Claw coverage from a short project list into a fuller family map, including ClawHub, Plugin Bundles, nix-openclaw, ResearchClaw Desktop App, and a clearer stack-layer taxonomy
- 2026-W16: Added a dedicated learning / RL / self-evolving layer to the guide, including Agent Lightning, Agent0, AgentEvolver, EvoAgentX, Acontext, and Awesome-Self-Evolving-Agents
- 2026-W14: Added 2026 Q1 signals for OpenClaw platformization, FutureHouse / Robin / BixBench, and Edison Scientific / Kosmos; refreshed ecosystem framing across the guide
- 2026-W14: Added recent Claude Code ecosystem signals, including anthropics/skills, wshobson/agents, SuperClaude, claude-code-router, Claude Squad, claude-task-master, and Repomix
- 2026-W13: Added a new plugin / bridge / registry layer to the guide, including cc-connect, OpenAlex Research MCP, Academia MCP, PapersWithCode MCP, and more Claw ecosystem positioning
- 2026-W13: Added core tools & platforms (InnoClaw, ResearchClaw, FARS, Orchestra, OpenClaw, EvoScientist); added Deep Research tools, OpenAI Prism, MCP Servers; switched all content to English; expanded to 35+ papers across 9 topic files
- 2026-W12: Redesigned README into a stronger landing page with cleaner hierarchy, card-style path selection, and a more visual ecosystem map
- 2026-W12: Hub-and-spoke architecture reorganization
- 2026-W12: Initial public release
Full history: CHANGELOG.md
MIT License ·
ai-science
literature-review
llm-agents
scientific-discovery
vibe-research
Similar repos
Significant stargazers
Seungwoo hong
98 followers · starred May 2026