Below is a high-level diagram of how data flows:

Acolyt is split into two major components:

​
Oracle API (Back-End)

  • Purpose: A foundational models to find similarities and trends between AI agents tokens and time-series data serving insights to the users.
  • Role: Ingests market data, trains ML models, and serves structured outputs (embeddings, forecasts, similarity rankings).

​
Conversational Agent (Interface)

  • Purpose: A framework interface built on an LLM framework (e.g., AI16z Eliza, Arc).
  • Role: Interprets user queries in natural language, calls Oracle API endpoints under the hood, and returns responses in human-readable format.

By separating the Oracle API (focused on robust ML and data pipelines) from the Conversational Agent (focused on user interaction), Acolyt’s architecture remains modular and scalable:

Phase 1 quickly delivers value by focusing on time-series-based analysis, which includes collecting high-frequency market data (such as price, volume, holders, and whales) and generating embeddings and forecasts with a foundational time series model. This enables queries like:

“What’s the outlook for agent X?”, “Which agents are similar to Y?” and “What good low cap agents should I invest in?”

Phase 2 adds graph-based modeling—incorporating social, team, whitepaper, holder, and other relationships to produce richer embeddings and more holistic insights into AI agents, thereby improving both forecast accuracy and token similarity analysis. This enables queries like:

“Give all the relevant data on agent X?”, “Give me a low cap with a good whitepaper” and “Which new agents has the most influential followers?”, “What the probability of agent X being a scam?”

Phase 3 [REDACTED]

The final result is a flexible, extensible system where ML-driven insights flow from the Oracle’s specialized models into natural-language explanations served by the Agent. This approach positions Acolyt to adapt continuously to new data types (social, dev, or on-chain) and user demands in the fast-evolving crypto landscape.

Want to know more about the technology? Check out our research & technology section.

OverviewRoadmap