luxas
An autonomous research colleague — from a question to a compiled manuscript, while you sleep.
Luxas is an open-source autonomous research agent system that transforms a research question into a complete LaTeX manuscript with citations, figures, and experiments. The multi-agent harness orchestrates literature surveys across OpenAlex, arXiv, and CrossRef, designs and executes computational experiments with blind implementation and testing, and generates publication-grade figures before compiling everything into a PDF report. Built on pi-mono and supporting multiple LLM backends (Claude, DeepSeek, OpenAI o3, GLM), it features crash recovery, file-backed state, and adversarial self-review loops. The system runs unattended for hours, using deterministic finish-gates and independent reviewer agents to ensure quality before completion.
Key Features
Use Cases
- 01Generating comprehensive literature surveys on specialized scientific topics with properly formatted citations
- 02Conducting computational research studies that combine theory, numerical experiments, and publication-quality figures
- 03Prototyping research workflows that require multi-hour unattended execution with crash recovery
- 04Producing reproducible research reports where every numerical claim traces to raw data via provref number-provenance
- 05Exploring cost-efficient research automation by redirecting expensive Claude models to DeepSeek (10× cheaper)
- 06Validating experimental code through adversarial blind implementation and testing by independent agents
Related Agents
View morehermes-agent
The agent that grows with you
agency-agents
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
openinterpreter
A coding agent for open models like Kimi K3
cline
Autonomous coding agent as an SDK, IDE extension, or CLI assistant.
luxas — FAQ
What is Luxas and what does it produce?+
Luxas is an autonomous multi-agent research system that takes a research topic and produces a complete compiled LaTeX PDF with literature citations, self-generated figures, and experimental results. It orchestrates 14 specialized agents to survey papers, design experiments, write code, audit figures, and iteratively refine the manuscript through adversarial review.
How do I install and run Luxas?+
Clone the repository, run npm install && npm link, install system dependencies (LaTeX, Python, tmux, poppler), set your ANTHROPIC_API_KEY, then run luxas init ~/research/topic --prompt "your question" followed by luxas run ~/research/topic. The system requires Node.js 22+, a full LaTeX distribution, and Python with matplotlib and numpy.
Which AI models and API keys does Luxas require?+
By default, Luxas requires an ANTHROPIC_API_KEY and uses Claude Opus, Sonnet, and Haiku across different agent roles. You can redirect the entire family to DeepSeek-v4 or Kimi with DEEPSEEK_API_KEY or KIMI_API_KEY via --profile dual or --model flags. Optional keys include OPENAI_API_KEY for the o3 math agent, GEMINI_API_KEY for figure generation, and BRAVE_API_KEY for web search.
Is Luxas free to use?+
Luxas itself is open-source MIT-licensed software, but it calls commercial LLM APIs that incur costs. A full research run costs approximately $20–80 using Claude models or $2–10 using the --profile dual preset with DeepSeek. Actual costs are logged per-run in the project's usage.log file.
How does Luxas handle crashes during multi-hour research runs?+
Luxas is designed to be crash-recoverable with stateless file-backed state. Re-running luxas run on a crashed project replays from checkpoint.jsonl and log.jsonl, reconstructs brain accounting (cost, tokens, counters), and resumes execution. Detached sub-agent processes continue running and their results are recovered through orphan recovery on the next turn.
What makes Luxas different from frameworks like LangGraph or AutoGPT?+
Luxas is research-domain-specific with a compiled PDF as the deliverable, not a general-purpose agent framework. It features file-backed crash recovery, detached sub-agent processes with heartbeat monitoring, native multi-model support via environment variables, and deterministic finish-gates enforced by code rather than prompts. It also includes adversarial self-review at three layers and a blind implementation-test pattern to prevent self-validation failures.
How do I install luxas?+
Open the source repository on GitHub and follow its README. luxas is a agent — MCP Agents Market links you directly to the official repo.
Is luxas free?+
luxas is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.