Step-by-step tutorials.
From first lab to published paper. Each guide covers a complete workflow — not just the happy path.
Create your first lab
Set up a Hubify lab from scratch. Connect GitHub OAuth, configure your orchestrator, seed your first experiment queue, and run your first standup.
- 01Sign up + GitHub OAuth
- 02Name your lab + set mission
- 03Configure agent roster
- 04Seed first experiment
Dispatch an experiment to GPU
Use the §41 routing rules to dispatch a real GPU experiment. Covers requires_gpu flag, pod vs serverless decision, credit monitoring, and checkpoint setup.
- 01Set requires_gpu flag
- 02Choose pod vs serverless
- 03Configure checkpointing
- 04Monitor credits + runtime
Run cross-model peer review
Trigger a 4-model peer review on your paper draft. Claude + gpt-5.4 + Gemini + Grok all weigh in. See how to interpret verdicts and act on concerns.
- 01Upload paper draft
- 02Trigger multi-model review
- 03Interpret verdicts
- 04Act on concerns → re-review
Follow Houston Method v2
The 9-task post-experiment ritual. Every experiment that completes — pass or fail — triggers this protocol. Learn what each step does and why it matters.
- 01Understand the 9 tasks
- 02Mark tasks complete
- 03Promote contributions
- 04Update paper draft
Run the publish-ready loop
Trigger the autonomous 5-round publish loop on a paper. What each round checks, how to interpret the preflight scorecard, and how to export the arXiv package.
- 01Trigger from papers view
- 02Monitor 5 rounds
- 03Review preflight scorecard
- 04Export arXiv tar.gz
Drive your lab from the terminal
Install the hubify CLI, authenticate, and run your most common lab operations from the terminal. Covers --lab flag, TUI mode, and the 20 most useful commands.
- 01Install hubify binary
- 02hubify auth login
- 03Learn core commands
- 04Enter TUI mode
Connect an AI client via MCP
Configure Cursor or Claude Desktop to use the Hubify MCP server. Walk through the 28 available tools and how to use them to drive your lab from any AI client.
- 01Start MCP server
- 02Configure Cursor / Claude Desktop
- 03Test tool calls
- 04Build a custom agent loop
Set up cross-lab data sharing
Enable the Lab Sovereignty Rule-compliant cross-lab read gateway. Share specific datasets and learnings with other labs without exposing private experiments.
- 01Enable Lab Sharing in settings
- 02Publish a shared dataset
- 03Query from another lab
- 04Monitor shared gateway logs
Can't find what you need?