| name | aris-infra |
| description | ARIS (Auto-claude-code-research-in-sleep) infrastructure setup and configuration.
Configures MCP servers for cross-model adversarial review, installs Python tools,
and validates environment. Run this first before using any other ARIS skills.
Use when: setting up ARIS, configuring review servers, "aris setup", "้
็ฝฎARIS".
|
| license | MIT |
| metadata | {"author":"wanshuiyin/ARIS","version":"1.0.0","repository":"https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep"} |
| allowed-tools | Bash, Read, Write, Edit, Glob, Grep |
ARIS Infrastructure Setup
Quick Start (One Command)
bash skills/aris-infra/setup.sh
This interactive script will: check prerequisites โ install dependencies โ register skills โ configure MCP reviewer server.
Manual Setup (if you prefer)
Overview
ARIS uses cross-model adversarial review โ Claude Code executes research tasks while an external LLM (GPT-5.4, Gemini, or others) provides critical review. This avoids the "self-play blind spot" where a single model reviewing its own work produces predictable feedback.
Prerequisites
- Python 3.10+
- Claude Code CLI
- At least one external LLM API key (OpenAI, Google Gemini, or MiniMax)
Step 1: Register MCP Servers
ARIS provides 5 MCP servers. Register the ones you need:
Core: Codex (GPT-5.4 Reviewer) โ Recommended
npm install -g @openai/codex
claude mcp add codex -s user -- codex mcp-server
Configure in ~/.codex/config.toml:
model = "gpt-5.4"
Alternative: Generic LLM Chat (Any OpenAI-compatible API)
claude mcp add llm-chat -s user -- python skills/aris-infra/mcp-servers/llm-chat/server.py
Environment variables:
LLM_API_KEY โ API key
LLM_BASE_URL โ API base URL (e.g., https://api.openai.com/v1)
LLM_MODEL โ Model name (e.g., gpt-4o)
LLM_FALLBACK_MODEL โ Fallback model on 504 errors
Alternative: Gemini Review
claude mcp add gemini-review -s user -- python skills/aris-infra/mcp-servers/gemini-review/server.py
Environment variables:
GEMINI_API_KEY or GOOGLE_API_KEY โ Google AI API key
GEMINI_REVIEW_MODEL โ Model (default: gemini-2.5-pro)
Alternative: Claude Review (Cross-session)
claude mcp add claude-review -s user -- python skills/aris-infra/mcp-servers/claude-review/server.py
Uses the claude CLI binary for reviews in a separate session.
Optional: MiniMax Chat
claude mcp add minimax-chat -s user -- python skills/aris-infra/mcp-servers/minimax-chat/server.py
Environment variables:
MINIMAX_API_KEY โ MiniMax API key
MINIMAX_MODEL โ Model (default: MiniMax-M2.7)
Optional: Feishu/Lark Notifications
claude mcp add feishu-bridge -s user -- python skills/aris-infra/mcp-servers/feishu-bridge/server.py
Environment variables:
FEISHU_APP_ID, FEISHU_APP_SECRET, FEISHU_USER_ID
BRIDGE_PORT โ HTTP server port (default: 9100)
Step 2: Install Python Dependencies
pip install httpx arxiv requests
Step 3: Verify Setup
claude mcp list
Available Workflows
After setup, use these one-click workflow skills:
| Skill | Command | Description |
|---|
aris-idea-discovery | /aris-idea-discovery | Full idea pipeline: literature โ ideas โ novelty โ review โ refine |
aris-experiment-bridge | /aris-experiment-bridge | Implement experiments, deploy to GPU, collect results |
aris-auto-review-loop | /aris-auto-review-loop | Multi-round cross-model adversarial review |
aris-paper-writing | /aris-paper-writing | Plan โ figures โ write LaTeX โ compile โ improve |
aris-rebuttal | /aris-rebuttal | Parse reviews โ strategy โ draft โ stress test |
aris-research-pipeline | /aris-research-pipeline | End-to-end: idea โ experiments โ review โ paper |
Bundled Resources
MCP Servers (mcp-servers/)
llm-chat/server.py โ Generic OpenAI-compatible bridge
gemini-review/server.py โ Gemini review with async jobs
claude-review/server.py โ Claude Code CLI review bridge
minimax-chat/server.py โ MiniMax-specific bridge
feishu-bridge/server.py โ Feishu/Lark notification bridge
Python Tools (tools/)
arxiv_fetch.py โ arXiv search and PDF download
semantic_scholar_fetch.py โ Semantic Scholar search with filters
research_wiki.py โ Persistent research knowledge base
watchdog.py โ GPU training/download monitoring daemon
Templates (templates/)
RESEARCH_BRIEF_TEMPLATE.md โ Research direction input
RESEARCH_CONTRACT_TEMPLATE.md โ Active idea working document
EXPERIMENT_PLAN_TEMPLATE.md โ Claim-driven experiment roadmap
EXPERIMENT_LOG_TEMPLATE.md โ Structured experiment results
NARRATIVE_REPORT_TEMPLATE.md โ Paper writing input
PAPER_PLAN_TEMPLATE.md โ Claims-evidence matrix
IDEA_CANDIDATES_TEMPLATE.md โ Compact top ideas
FINDINGS_TEMPLATE.md โ Cross-stage discovery log
Troubleshooting
- MCP server not found: Ensure
claude mcp add was run with -s user flag
- API key errors: Set environment variables in your shell profile (~/.zshrc or ~/.bashrc)
- Python import errors: Run
pip install httpx arxiv requests
- Codex not installed: Run
npm install -g @openai/codex