Cairn

A free AGPLv3 coordination layer that lets multiple AI agents work together without burning the token budget, by coordinating through a passive shared blackboard and minimal per-agent context instead of conversation. Existing tooling is reactive (budget caps, observability); Cairn is preventive: it removes the unbounded-conversation structure that causes context and cost to explode. Measured on a controlled A/B (identical task, same model, only the context architecture changed): 54% fewer input tokens, 65% lower cost. The benchmark harness is published so the saving is reproducible. Framework-agnostic; composes with LangGraph and CrewAI. Also submitted to NLnet NGI Zero Commons (code 2026-06-385); FLOSS/fund support would fund maintenance, framework adapters, and the public benchmark.