Vibe Research Guide
From core Vibe Research platforms to supporting tools — find the right combination for your workflow.
Core Vibe Research Platforms Agent systems and platforms purpose-built for automated scientific research:
Project
Focus
Link
Highlights
InnoClaw AI research innovation agent
GitHub Open-source, literature analysis to idea generation
ResearchClaw Personal research assistant
GitHub Open-source, CLI/Web/Slack, integrated research/data/general tools
FARS (Analemma) Fully automated research system
GitHub · Website Fudan MOSS team, 4 modules end-to-end (Ideation→Planning→Experiment→Writing)
AI-Scientist End-to-end research automation
GitHub Sakana AI, idea→experiment→paper→review pipeline
EvoScientist Self-evolving AI scientist
GitHub Multi-agent with persistent memory, outperforms 7 SOTA systems
Orchestra AI-for-Science platform
Website Cloud-based, no local deployment needed
Agent Laboratory Human-AI collaborative research
GitHub Human-in-the-loop, researcher stays involved throughout
FutureHouse Platform Scientific agent platform
Website Public platform for specialized science agents and API-style workflows
Edison Scientific AI scientist platform for R&D teams
Website · Kosmos Commercial scientific-discovery platform with private-data and enterprise workflow focus
Related Claw Ecosystem Projects & Infrastructure
Project
Focus
Link
Highlights
OpenClaw General-purpose personal AI assistant framework
GitHub · Docs Cross-platform assistant with an extensible skill ecosystem
ClawHub Public registry for OpenClaw skills and plugins
GitHub · Docs Search, install, publish, and version OpenClaw capabilities through a dedicated ecosystem layer
OpenClaw Plugin Bundles Compatibility layer for third-party bundle ecosystems
Docs Reuses Codex, Claude, and Cursor bundle formats inside OpenClaw-native features
nix-openclaw Declarative Nix packaging and deployment
GitHub Reproducible setup, upgrades, and rollback path for OpenClaw environments
ScienceClaw Self-evolving scientific research agent
GitHub Research-oriented Claw variant with persistent memory and evolving skills
ScienceClaw (alt repo) Alternative research-lab implementation
GitHub Research workflow automation with broad database/tool coverage
MetaClaw Self-evolving agent framework
GitHub · Website Online learning framework for continuously improving Claw-style agents
AutoResearchClaw Autonomous research pipeline
GitHub End-to-end academic pipeline from topic to paper with code, review, and reports
ResearchClaw Desktop App Desktop paper reading and note-taking workflow
Website PDF chat, note-taking, agentic search, local-first workflow
awesome-openclaw-skills Community skill directory
GitHub Fast discovery layer for OpenClaw-compatible skills, examples, and ecosystem patterns
More ecosystem positioning: → Claw Park
Dimension
InnoClaw
ResearchClaw
FARS
AI-Scientist
Orchestra
Open-source
✅
✅
✅
✅
❌ (cloud)
Coverage
Literature→Idea
Literature→Data→Experiment
Idea→Paper (full)
Idea→Paper (full)
Full pipeline
Interface
CLI
CLI/Web/Slack
Web
CLI
Web
Difficulty
⭐⭐
⭐⭐
⭐⭐⭐
⭐⭐⭐
⭐
Best for
Beginners / Researchers
Individual researchers
Full automation
Full automation
Zero-code users
OpenClaw is an open AI agent skill ecosystem spanning assistant, research, search, and workflow capabilities, providing extensible skill support for the platforms above.
The strongest current signal is that OpenClaw is no longer just a single assistant shell:
It has a public skill-discovery layer through ClawHub .
It is pushing cross-ecosystem installability through Plugin Bundles , including compatibility paths for other coding-agent ecosystems.
It now has a clearer packaging path through nix-openclaw , which matters for reproducible self-hosting and operations.
Its value is increasingly in acting as a gateway, distribution layer, control surface, and deployment substrate for broader agent workflows.
If you are confused by the growing number of Claw-branded projects, the cleanest mental model is:
Layer
Representative project
Role
Foundation
OpenClaw
Base runtime, gateway, and control layer
Registry / discovery
ClawHub · awesome-openclaw-skills
Helps users find, install, compare, and publish skills and plugins
Compatibility layer
OpenClaw Plugin Bundles
Imports adjacent coding-agent bundle ecosystems into OpenClaw
Deployment layer
nix-openclaw
Adds reproducible packaging, installation, and rollback
Research workspace
InnoClaw · ResearchClaw Desktop App
Grounded surfaces for files, papers, notes, and local reading
Personal research copilot
ResearchClaw
Day-to-day research assistance across search, notes, and writing
Scientific specialist
ScienceClaw
Research-oriented, self-evolving scientific collaborator
Evolution engine
MetaClaw
Learns skills from real conversations and improves over time
Autonomous pipeline
AutoResearchClaw
Turns an idea into a full paper pipeline with minimal human intervention
Full guide: → Claw Park
Plugins, Bridges & Registries The stack is no longer just "pick an agent." A new layer of bridges, plugin surfaces, and registries now determines how easily a research workflow can be extended and operated.
Resource
Link
What it adds
Why it matters
cc-connect GitHub Bridges Claude Code, Cursor, Gemini CLI, Codex, Goose, and more into chat apps
Useful when teams want to run coding or research agents from Feishu/Lark, Slack, Telegram, or WeCom instead of a local terminal
ClawHub Docs Public discovery surface for OpenClaw-compatible skills and tools
Shows OpenClaw moving toward a real skill-distribution layer rather than a closed assistant
OpenClaw Plugin Bundles Docs Compatible bundles for importing capabilities across agent ecosystems
Useful for understanding how agent ecosystems may converge through bundle compatibility instead of isolated plugin silos
Claude Code extensibility Docs Official hooks, slash commands, subagents, and MCP integration surfaces
Defines the plugin/customization model around one of the most common terminal agents
awesome-claude-code-plugins GitHub Curated list of Claude Code prompts, commands, hooks, subagents, and MCP servers
Fastest way to see how the Claude Code plugin layer is evolving in practice
MCP Registry Website Official discovery and installation surface for MCP servers
Reduces friction when testing or installing new tool integrations
awesome-mcp-servers GitHub Broad discovery map of the MCP ecosystem
One of the highest-signal indexes for scouting connectors quickly
awesome-openclaw-skills GitHub Community-curated OpenClaw skill directory
Shows the skill-distribution layer growing around the Claw ecosystem
Literature Discovery & Search
Tool
Link
Highlights
Cost
Semantic Scholar
website Allen AI, AI-powered search, open API
Free
Connected Papers
website Citation graph visualization
Free
Research Rabbit
website Continuous paper recommendations based on your collection
Free
Litmaps
website Interactive literature maps
Free / Pro
Google Scholar
website Broadest coverage, supports Alerts
Free
Literature Reading & Synthesis
Tool
Link
Highlights
Cost
Elicit
website Enter a research question, auto-retrieve and extract data
Free
Consensus
website Evidence-based answers from papers
Free
NotebookLM
website Google, upload papers for conversational reading
Free
PaperQA2
GitHub Open-source RAG literature Q&A with precise citations
Free
OpenScholar
GitHub Allen AI, large-scale literature synthesis
Free
Perplexity
website AI search engine with citations
Free
Tool
Link
Highlights
Cost
OpenAI Deep Research
website Multi-step web research, structured reports, and cited synthesis
ChatGPT
Gemini Deep Research
website Google research copilot for broad web synthesis and planning
Gemini
Perplexity Deep Research
website Free, academic focus mode, multi-source synthesis
Free
Prism
website Scientific writing workspace for paper drafting, citations, and LaTeX-heavy workflows
Free
AI Scientist Platforms & Current Signals One of the clearest 2026 shifts is that AI-scientist systems are no longer only papers or repos. They are becoming public platforms, products, and applied benchmarks.
Signal
Link
Why it matters
FutureHouse Platform Website Shows the public-platform direction for specialized scientific agents, not just one-off demos
Robin Announcement High-signal end-to-end demonstration of multi-agent scientific discovery and validated wet-lab-style workflow claims
BixBench Announcement Signals that more realistic, domain-specific benchmarks are becoming central to evaluating research agents
Edison Scientific / Kosmos Website · Kosmos Shows how AI-scientist ideas are being packaged into persistent scientific-discovery platforms for real R&D teams
Learning, RL & Self-Evolving Agents The stack is no longer only "tool-using agents." A distinct learning layer is forming around training, online improvement, skill extraction, and persistent memory.
Resource
Link
What it does
Why it matters
Agent Lightning GitHub Trains arbitrary agents with RL, automatic prompt optimization, and SFT with minimal code changes
Strongest current signal that agent training is becoming framework-agnostic infrastructure
Agent0 GitHub Zero-data self-evolving language and vision-language agents through tool-integrated reasoning
Useful reference for agents that generate their own training signal instead of relying on curated datasets
AgentEvolver GitHub End-to-end self-evolving training system with self-questioning, self-navigating, and self-attributing
Good mental model for how exploration, experience reuse, and credit assignment can be unified
EvoAgentX GitHub Framework for building, evaluating, and evolving agentic workflows
Pushes the field from single-prompt tuning toward workflow-level evolution
Acontext GitHub Context and memory platform that turns sessions into persistent skills and reusable operating knowledge
Shows how memory and skill accumulation are becoming part of the agent-improvement stack
Awesome-Self-Evolving-Agents GitHub Survey-style map of self-evolving agents and optimization methods
Best compact overview when you want the broader literature and repo landscape, not only one framework
Research MCP Servers & Academic Connectors Model Context Protocol (MCP) servers that integrate academic search into AI workflows:
Server
Link
Highlights
Semantic Scholar MCP
GitHub Search papers, analyze citations, paper recommendations
OpenAlex Research MCP
GitHub OpenAlex-powered literature search, citation graph exploration, and research trend analysis
Academia MCP
GitHub Aggregates arXiv, ACL Anthology, Semantic Scholar, and academic search into one MCP surface
PapersWithCode MCP
GitHub Bridges papers to code repos, datasets, tasks, methods, and leaderboard context
arXiv MCP Server
GitHub Search & retrieve arXiv papers with filtering by author, category, date
PubMed MCP Server
PulseMCP Biomedical literature search and retrieval
Tool
Link
Highlights
Cost
Writefull
website AI proofreading designed for academic writing
Free / Pro
Paperpal
website Backed by Springer Nature
Free / Pro
Grammarly
website General writing assistant with academic mode
Free
OpenAI Prism
website OpenAI for Science — ChatGPT embedded in a scientific text editor
Free
Research agents increasingly fail or succeed based on their coding execution layer : can they inspect repos, edit files safely, run experiments, and iterate on failures?
Tool
Link
Highlights
Cost
Claude Code
Docs Terminal-native coding agent with strong repo workflows, MCP support, and approval controls
Usage-based
Codex
GitHub OpenAI coding agent for local and cloud task execution
Usage-based
Cursor Background Agents
Docs Async remote agents for longer engineering tasks and parallel execution
Pro / Teams
GitHub Copilot Coding Agent
Docs Issue-to-PR workflow directly inside GitHub
Paid GitHub plan
Gemini CLI
GitHub Open-source terminal agent with Google Search grounding and MCP support
Free / usage-based
OpenHands
GitHub Open-source full-stack agent platform with sandboxed execution
Free
More workflow detail: → Vibe Coding
Tool
Link
Highlights
Cost
Cursor
website AI-native code editor
Free / Pro
GitHub Copilot
website AI coding assistant
$10/mo
OpenHands
GitHub Open-source full-stack agent
Free
SWE-agent
GitHub Code repair agent
Free
Weights & Biases
website ML experiment tracking
Free (personal)
Tool
Link
Highlights
Cost
Zotero
website Open-source, rich ecosystem (AI plugins available)
Free
Zotero + ZoteroGPT
GitHub GPT plugin for Zotero
Free
Paperpile
website Deep Google Docs integration
$3/mo
Adjacent Agentic Workflows Not every useful "vibe" workflow is research. App prototyping, design iteration, slide generation, and knowledge work should stay in a separate lane so this repo does not blur core scientific automation with general-purpose creation.
Tool
Link
Highlights
Cost
v0
website Prompt-to-UI / app generation with code export and iteration
Free / usage-based
Lovable
website Product-builder workflow from prompt to deployed app with GitHub sync
Free / paid
Replit Agent
Docs In-browser agent for app building, debugging, and deployment
Free / paid
Figma AI
website Design ideation, Make-style prototyping, and interface generation
Paid
NotebookLM
website Source-grounded workspace for briefs, notes, and content synthesis
Free / paid
Broader framing: → Vibe Anything
What you want to do
Recommended
Quick Vibe Research demo Elicit + NotebookLM (zero-code)
Deploy a personal Research Agent InnoClaw or ResearchClaw
Need a public science-agent platform FutureHouse Platform
Need a commercial AI-scientist platform for team workflows Edison Scientific / Kosmos
Operate agents from team chat cc-connect
Scout new connectors quickly MCP Registry or awesome-mcp-servers
Literature RAG Q&A PaperQA2
Need literature plus benchmark/code metadata OpenAlex Research MCP + PapersWithCode MCP
Need agent training / optimization Agent Lightning
Need self-evolving agent infrastructure Agent0 or AgentEvolver
Need persistent skill / context memory Acontext
Full automated research FARS or AI-Scientist
Cloud, no deployment Orchestra
Complete research workflow ResearchClaw + Zotero + Cursor + W&B
Researcher who also codes Deep Research + Claude Code or Codex + W&B
Async issue-to-PR engineering Cursor Background Agents or GitHub Copilot Coding Agent
Prompt-to-app prototyping v0 or Lovable
Home : README · Prev : Getting Started · Next : Surveys