npx claudepluginhub ryansinger/syndicateSpins up a self-governing outfit that iterates on its own work. Give it a goal and it evolves its approach through attempts, self-evaluation, and skill refinement. Supports job mode (converge and ship) and venture mode (ship, discover improvements, repeat). A coherence agent prevents runaway complexity. Learned agents and domain skills accumulate as the syndicate gains experience.
A Claude Code plugin that spins up a self-governing outfit to do a job. Give it a goal (build an app, write a contract, design a system) and it stands up an organization that attempts the work, scores itself, evolves its approach, and ships when it's done.
/plugin marketplace add RyanSinger/syndicate
/plugin install syndicate@RyanSinger-syndicate
/reload-plugins
Just tell Claude what you want:
"Build me an invoice tracking app"
"Write a partnership agreement for a joint venture between X and Y"
"Design a REST API for a booking system"
Two modes:
Just describe what you want. The syndicate figures out which mode fits, or you can say "venture" to keep it going.
Three roles, same pattern as organizational governance:
As the syndicate learns, it promotes recurring patterns into learned agents (specialized subagents) and domain skills (reusable knowledge imported from plugins or distilled from experience).
The coherence agent never sees code. Only scores, complexity metrics, and commit messages. Its prompt is fixed and can't be modified by the system. This structural separation, borrowed from TurkoMatic (2011), is what keeps the syndicate from spiraling.
The system governs itself. It decides when to upgrade models, when to revise its own success criteria, and when to stop. No generation count. No hand-holding.
Combines Meta's HyperAgents (2026) with failure-mode research from TurkoMatic (Kulkarni et al., 2011). See plugin/skills/run/references/architecture.md.
MIT