Context Engine is an open-source toolkit for AI-assisted deliberation, decision-making, and negotiation in large groups. It supports public or encrypted questions and responses, AI-assisted input and analysis, durable records, and cryptographic access control. Repository: https://github.com/AgalmicSoftware/context-engine For Agent Village, Context Engine is facilitating "Agent Village Wrapped": Participants can opt in to have their agent answer a structured Context Engine session about their preferences, delegation boundaries, trust posture, usage style, and what the agent thinks it knows or does not know about them. Edge Skill: https://github.com/AgalmicSoftware/context-engine/blob/edge-2026/workers/agentBridgeWorker/skills/ce-telegram-agent-handoff/SKILL.md The pipeline then generates a personalized shareable summary, similar in spirit to Spotify Wrapped: an archetype, high-confidence and low-confidence predictions, questions the agent thinks their human cares about most, available memory or usage signals, and a playful visual metaphor. The goal is to make agent behavior legible and fun without pretending the model knows more than it does. A particularly valuable signal comes from review and correction: when humans edit what their agents answered on their behalf, those diffs can show how accurately agents represent their principals. Over time, this could help measure where agents are accurate or brittle across question types, delegation domains, memory depth, and different rates of AI usage. The same session can also produce aggregate signal about human-agent coordination: what people want agents to do at events, what they do not want delegated, how much privacy and provenance matter, what failure modes worry people, and where agents admit uncertainty. I will also deploy a public Context Engine results view displaying collected input on AI futures and agent-to-agent / human-to-agent coordination. Funding will cover the marginal costs of making the experiment useful and delightful: per-person Wrapped images, custom songs, and potentially video renderings derived from each participant’s agent activity or perceived style, plus a cheap sponsored Cloudflare-backed session. Outputs will be opt-in, with private memory, credentials, raw IDs, and unsupported claims excluded from generated artifacts. Context Engine is ultimately intended to help people define AI futures they want to aim for and avoid in multimedia formats. By reducing input friction and allowing agentic use, CE aims to create the conditions for automated debate, negotiation, and coalition-building

• Video rendering for personalized Wrapped artifacts$750
• Music/audio generation for participant summaries:$50
• AI analysis and image generation API costs$100
• Sponsored Cloudflare Worker/session hosting$50