</>MCP Agents Market
Agent

DeepScientist

by ResearAI3.3kTypeScriptUpdated 2026-06-28

Now, Stronger AI Pushes Frontiers, Stronger Our Shared Future.

DeepScientist is a local-first autonomous research agent that orchestrates end-to-end scientific workflows from baseline reproduction through experimental iteration to paper-ready outputs. Built around Findings Memory, Bayesian optimization, and a Research Map, it transforms research tasks into persistent Git repositories that accumulate knowledge across experiment rounds. The agent supports multiple AI backends (Codex, Claude Code, Kimi Code, OpenCode) and provides web, terminal, and messaging interfaces for monitoring and human-in-the-loop intervention. Designed for graduate students and research teams, it handles environment setup, dependency resolution, experiment branching, result analysis, and LaTeX compilation within a unified workspace.

Key Features

End-to-end autonomous research workflow from paper ingestion to baseline reproduction, experiment iteration, and publication drafts
Local-first architecture with one Git repository per research quest, preserving all branches, failed paths, and successful routes
Findings Memory system that accumulates knowledge across experiment rounds and informs Bayesian optimization for next hypotheses
Research Map visualization showing experiment branches, baselines, and accumulated knowledge structure
Multi-runner support for Codex, Claude Code, Kimi Code, and OpenCode with fallback to bundled helpers
Human-in-the-loop design allowing pause, takeover, plan editing, and continuation at any point
Multiple interaction surfaces: web workspace (default port 20999), terminal TUI, and messaging connectors (WeChat, QQ, Telegram, WhatsApp, Feishu)
Integrated paper authoring with PDF/LaTeX compilation, figure generation, and experiment result organization

Use Cases

  • 01Reproducing machine learning baselines from papers with automatic environment and dependency resolution
  • 02Running long-horizon experiment loops with automated ablations, comparisons, and hypothesis generation
  • 03Managing multi-branch research exploration where failed routes are preserved for future reference
  • 04Collaborative research projects where team members monitor progress through web or messaging interfaces
  • 05Generating publication-ready materials including figures, result tables, and LaTeX drafts from experiment outputs
  • 06Server-based research workflows controlled via terminal TUI while monitoring progress remotely

Related Agents

View more

DeepScientist — FAQ

What is DeepScientist?+

DeepScientist is an autonomous AI research agent that runs locally and manages the complete scientific research lifecycle, from reading papers and reproducing baselines to running experiments and generating publication drafts. Unlike one-shot tools, it maintains persistent state across long research projects using Git repositories and accumulated memory.

How do I install DeepScientist?+

Install globally via npm with 'npm install -g @researai/deepscientist', ensure you have an authenticated runner (codex, claude, kimi, or opencode), then run 'ds --here' to launch. Alternatively, clone the GitHub repository and run 'bash install.sh'. The system requires Node.js, npm, git, and Python 3.11+.

Which AI models and clients does DeepScientist work with?+

DeepScientist supports four built-in runners: Codex, Claude Code, Kimi Code, and OpenCode. It can also connect to Gemini via OpenCode and Ollama via any runner with local model backend configuration. The agent provides web, TUI, and messaging connector interfaces rather than integrating as a plugin into other clients.

Do I need API keys or paid accounts?+

Yes, you need a working installation of at least one runner (codex, claude, kimi, or opencode) with valid authentication. Each runner requires its own API credentials or account. DeepScientist itself is open-source under Apache 2.0, but the underlying AI models may have their own costs.

Is DeepScientist free to use?+

DeepScientist is free and open-source software released under the Apache 2.0 license. However, running it requires API access to one of the supported AI backends (Codex, Claude, Kimi, OpenCode), which may have associated costs depending on the provider.

What platforms does DeepScientist support?+

DeepScientist fully supports Linux and macOS. Native Windows support is experimental; Windows users are strongly recommended to use WSL2. The system runs locally by default, with all code, experiments, and research state stored on your own machine or server.

How do I install DeepScientist?+

Open the source repository on GitHub and follow its README. DeepScientist is a agent — MCP Agents Market links you directly to the official repo.

Is DeepScientist free?+

DeepScientist is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches