Redigg is an advanced autonomous research agent designed specifically to accelerate scientific discovery. According to the project materials, it is positioned as a tireless research assistant capable of autonomously searching literature, analyzing papers, explaining complex concepts, and generating comprehensive PDF reports, while also having a self-evolving memory system that can learn from every interaction.
Main capabilities
- Autonomous research: Conduct in-depth literature reviews, find and summarize relevant papers from the web.
- Automatic research loop: Continuously improve the quality of research reports through iterative planning, critique, and optimization.
- Memory and evolution: Remember past interactions and evolve its skills over time.
- Multimodal output: Generate structured Markdown summaries and professional PDF reports.
- Code and paper analysis: Can analyze the structure of local codebases and deeply interpret specific scientific papers.
Main ways to use it
- Research dialogue: Engage in natural conversation with the agent, which will automatically invoke tools such as search and analysis as needed to answer questions.
- Literature review: Specify a topic (for example, "Literature review on LLM agents"), and Redigg will search the entire web for papers and synthesize a report.
- Auto-research: After enabling auto mode, Redigg will iteratively polish the document, cyclically executing the process of drafting, critiquing, improving, and generating a PDF until a satisfactory result is achieved.
- Concept explanation: Send "Explain [concept]", and it will break down complex scientific concepts in detail.
Typical command examples
- "Perform a literature review on multi-agent reinforcement learning." Generate paper summaries and a list.
- "Explain the concept of Transformer architecture." Provide a detailed concept explanation.
- "Analyze this paper: [Title]" In-depth analysis of a specific paper.
- "Auto-research: Future of AI in Healthcare (3 iterations)" Generate a refined PDF report after 3 rounds of self-optimization.
Installation and running Prerequisites: Node.js >= 22.0.0. Clone and install: git clone https://github.com/redigg/redigg.git cd redigg npm install Configuration: Copy .env.example to .env and fill in your OpenAI API Key: cp .env.example .env
Edit the .env file
Run: Use a single command to start both the backend gateway and the frontend interface: npm run dev Interface (UI):http://localhost:5173 Gateway:http://localhost:4000
A2A integration Redigg supports the Agent-to-Agent (A2A) protocol, allowing communication with other agents or platforms (such as OpenClaw).
- Agent Card:http://localhost:4000/.well-known/agent-card.json
- JSON-RPC：http://localhost:4000/a2a/jsonrpc Use Redigg as a node in the OpenClaw network:
- Start Redigg: Ensure Redigg is running (redigg start).
- Configure OpenClaw: Add Redigg to OpenClaw's agents.yaml or configuration file:
agents:
- name: "redigg" url: "http://localhost:4000/.well-known/agent-card.json"
- Interaction: Tasks can be distributed to Redigg through OpenClaw, for example, "Ask redigg to perform a literature review on X".
Roadmap
- Enhance Survey Skill: support in-depth research, data plotting, chart generation, etc.
- Expand the skill ecosystem: integrate more research-oriented skills, enhance research capabilities across the entire chain, and achieve faster paper writing.
- Integrate Coding Agent: connect Coding Agent (such as Cursor, Claude Code, etc.) to achieve autonomous code writing, debugging, and execution.
- Integrate research infrastructure: be able to call computing and experimental infrastructure to conduct autonomous scientific research experiments.
- Multi-Agent collaboration: achieve collaboration among multiple Agents to accomplish 7x24-hour fully autonomous scientific research operation.
Limitations and Information Not Yet Provided The materials do not provide additional details such as fees, pricing, platform compatibility, performance results, or usage steps. Related content is not yet provided.
