deep-research-agent

DeepSeek Harness agent in a locked-down Apple container: default-deny egress via pf, self-hosted web research, loop guard

AutoMate is a Personal Autonomous reAct Agent built for seamless daily task management.

DRIA V2 is a local-first Deep Research and Intelligence Agent. It combines a FastRTC voice UI, local llama.cpp-compatible reasoning, SearXNG search, self-hosted Firecrawl scraping, markdown agent skills, optional Anam avatar video, and background multi-agent research jobs into a Docker-ready personal AI workspace.

26-tool autonomous research agent — academic papers, arXiv, PubMed, PubChem, Census, web search. x402 micropayments, any LLM.

MiroThinker is an open source deep research agent optimized for research and prediction. It achieves a 80.8% Avg@8 score on the challenging GAIA benchmark.

The RigorQuest AI is a sophisticated multi-agent system designed to conduct comprehensive research on user-specified topics within a chosen domain. It leverages AI-driven tools, including the Meta LLaMA model, to generate targeted research questions, perform in-depth analysis, and compile findings into a professional report.

A multi-agent research system built with LangGraph and TypeScript that orchestrates AI agents to perform comprehensive research.

🔍 Artifact-agnostic audit / verify / red-team framework — portable across 🤖 Claude, Codex, Cursor, Gemini & Copilot (AGENTS.md) + 🪶 low-context Lite. 🌐 Researches current standards (Stage R), 🔁 Ralph-loops audit→fix→re-audit to a clean pass, 🧩 multi-vendor sub-agents. 🏛️ @schatt93 · 🤝 Claude as coding & research partner.

Automate research tasks with local LLMs, real-time web search, and structured report generation.

Legumista is an MCP server for Legume Information System projects

A sophisticated, asynchronous multi-agent AI system built with LangGraph, LangChain, and Tavily Search to autonomously scope, research, and synthesize comprehensive reports on any topic.

Turn Claude Code into a state-of-the-art deep research agent with a tier-adaptive 16-step pipeline.

Deep Research skill for OpenClaw/ClawHub — 26 research endpoints via x402 micropayments.

MiroRL is an MCP-first reinforcement learning framework for deep research agent.

可验证的Deep Research Agent Harness，支持多智能体协作、长程检索、证据溯源、引用核验、断点恢复、多模态输入与可复现评测。Verification-centric Deep Research Agent Harness with multi-agent orchestration, long-horizon retrieval, evidence provenance, citation auditing, checkpoint recovery, multimodal input, and reproducible evaluation.

Java/Spring AI Agent runtime inspired by DeerFlow, with Graph-first execution, Deep Research, MCP, Skills, Memory, SSE streaming, durable checkpoints, and sandboxed tools.

Build open source deep research agents using multiple models, search tools, and MCP servers.

Autonomous deep research engine designed to bypass standard web scraping by reverse-engineering backend APIs. Conducts 5-layer investigations, cross-verifies multi-source evidence, and delivers 1:1 fact-grounded synthesis with zero AI hallucinations.

Deep-Research