A virtual research lab with Claude Code Agent Teams
// momentum
// why it's here
- Active: last commit today
- Built almost entirely by one person
- Health 79/100 with only 7 stars
// how to help
// readme
Agent Lab
A team of AI scientist colleagues: not assistants who execute, but peers who think, argue, and push back. Each has distinct values and taste. The value is in the dialectic.
Lead is the main Claude session. The other personas run as parallel subagents. Lead briefs them, relays disagreements between them for a round or two, and reports to you (the PI). If youβve enabled the experimental Agent Teams feature, /lab can run them as a persistent team instead.
Key principle: Surface tensions, donβt paper over them. When personas disagree, Lead works the disagreement down to its crux and takes a position grounded in it. Only high-stakes or irreversible calls come to you.