Research Preview

MetaCRDT

A convergence substrate for structured coordination across distributed runtimes.

Why MetaCRDT?

Operational systems are edited by humans, services, workflows, and agents. Those edits are not just document changes. They are facts, tasks, reviews, violations, proposals, messages, and decisions that need to remain meaningful as they move across distributed runtimes.

MetaCRDT is a research substrate for making those structures converge. It combines CRDT-style merge semantics, append-only fact logs, typed coordination objects, and deterministic derivation. Together, these primitives let us express arbitrarily rich workflows, from compliance data rooms to generative group games, while preserving:

• Fact Convergence — Operational facts, tasks, documents, violations, and proposals merge across distributed runtimes.
• Provenance — Every assertion carries authorship, causality, replay history, and enough context to explain how state changed.
• Derived Coherence — Queries, constraints, violations, workflows, and views are recomputed from shared facts rather than copied between tools.
• Agent Participation — Agents observe facts, propose actions, and leave mergeable records under the same semantics as human operators.

Ontologies, data rooms, conversations, response surfaces, and collaborative worlds are not separate products in this model. They are elaborations over the same convergent substrate.

Research Elaborations

• ontology.metacrdt.com

  Operational ontologies for real-world systems: entities, facts, relationships, constraints, workflows, views, and agent proposals elaborated over convergent shared state.

• datarooms.metacrdt.com

  Structured coordination spaces for compliance, onboarding, due diligence, and regulated operations where every artifact, review, request, task, and decision is part of a mergeable fact log.

• groupchat.metacrdt.com

  Conversation systems that treat every message, reply, annotation, reaction, and agent intervention as first-class replicated state rather than an ephemeral chat log.

• prompt-response.metacrdt.com

  Prompt-native response surfaces where forms become conversations, submissions become facts, and agents can synthesize, extract, and route structured outcomes.

• threadquest.metacrdt.com

  Procedural narrative worlds in which an AI Dungeon Master orchestrates branching story graphs, persistent world state, and multiplayer interaction in shared chat.

Architectural Stack

• Append-Only Facts — Entity facts, task facts, document facts, and proposal facts are recorded as replayable operations with explicit merge semantics.
• Identity & Provenance — Actors, signatures, authorship, timestamps, and causal references make every change inspectable and attributable.
• Deterministic Derivation — Queries, constraints, violations, rules, and typed views derive coherent interpretations from shared state.
• Intent & Coordination — Tasks, workflows, approvals, and proposal objects express why state changes happen, not just the resulting mutations.
• Agent Participation — Agents analyze, synthesize, propose, and merge changes alongside human operators while preserving provenance.
• Transport & Execution — Distributed runtimes propagate shared state across browsers, servers, Durable Objects, background workers, and peer-to-peer sessions.

Participate

• Read the Docs — Deep-dive notes on fact models, merge semantics, derived views, and agent proposals.
• Join the Alpha Cohort — Access private registries and reference implementations.
• Propose a Lab — Submit a two-page abstract describing how convergent facts can unlock your domain.

MetaCRDT is an open, evolving substrate; each sub-domain is a living laboratory. If your work demands structured coordination under real-world distributed conditions, we invite you to build with us.

Join the Research Incubator