DreamGraph
The cognitive MCP server that powers NoFS research
What is DreamGraph?
DreamGraph is a cognitive graph engine built as a Model Context Protocol (MCP) server. It doesn't just store data — it thinks about it. Through dream cycles, causal reasoning, adversarial analysis, and metacognition, DreamGraph discovers structural insights that no static tool can find.
The NoFS hypothesis emerged directly from DreamGraph's own cognitive analysis of its codebase. The system observed that its graph already contained richer relationships than the file system could express, and proposed: what if we stopped pretending files are the source of truth?
How It Works
- Dream Cycles — Autonomous analysis runs that explore the graph, find patterns, generate hypotheses
- Causal Reasoning — Understands why code exists, not just what it does
- Adversarial Analysis — Actively tries to break its own theories to validate them
- Metacognition — Monitors its own reasoning quality and adjusts strategies
- Federation — Multiple DreamGraph instances can share learned patterns across boundaries
- Temporal Analysis — Tracks how code evolves over time and predicts trajectories
DreamGraph as NoFS Testbed
DreamGraph is both the research engine and the first test subject for NoFS. Its own codebase is the graph that the NoFS hypotheses are tested against. Every entity, every relationship, every dream insight is a data point in the research.
◈ Self-Referential Research
DreamGraph analyzes itself with cognitive tools, discovering patterns in its own structure. The NoFS hypothesis was born from this self-analysis.
◈ Live Graph Store
45+ entities, 100+ edges, architectural decisions, capabilities — all stored as graph data, not files. This is NoFS, running in production.
◈ 30+ MCP Tools
A rich toolkit for cognitive operations: dream cycles, causal queries, temporal insights, federation, and more. All interacting through the graph.
Explore DreamGraph
DreamGraph is open source on GitHub. Explore the code, run your own instance, contribute to the research.
View on GitHub → See the Research →