CoinStats builds AI agent for real-time, private crypto research

CoinStats is developing an AI research agent that pulls live on-chain, market and social data from more than 120 blockchains and offers a Private Mode and developer APIs, the company’s CEO Narek Gevorgyan said. The product will use CoinStats’ existing data layer to deliver real-time, multi-source crypto research and tools for developers.

CoinStats began as a portfolio tracker, and users asked for more than balance checks. “Tracking a portfolio is the easy part. The hard part is understanding what to do next,” Gevorgyan said, describing users who spent hours switching between social feeds, block explorers, exchanges and dashboards to make one decision.

The system uses specialized sub-agents that run in parallel and read live sources. One sub-agent pulls real-time news, another scans social sentiment, another reads on-chain transactions across 120+ chains, while others check exchange metrics or analyze a user’s portfolio. Those components report to a synthesizer that returns answers with interactive tables and charts rather than a long block of text.

CoinStats plans three answer modes to match different research needs: Deep Research for multi-source reports, Backtesting to validate strategies against historical data, and Fast Mode for quick lookups.

Privacy and result verification are part of the design. Private Mode routes queries through an encrypted, decentralized system from Venice AI so external AI providers do not see user data, Gevorgyan noted. The product will attach sources to claims and include backtesting tools so users can verify outputs against historical market behavior. CoinStats positions the agent as a research tool to support users’ own decision making rather than replace their judgment.

For developers, CoinStats will expose a normalized dataset through a Crypto API and an MCP server for AI agents and development environments. The company reports coverage of more than 100,000 coins, 200+ exchanges, 120+ blockchains and DeFi positions resolved across over 10,000 protocols at the wallet level. The APIs aim to let agents and tools access portfolio state, market data, news and on-chain context without building separate aggregation systems.

Gevorgyan outlined three near-term goals for the project: expand an industry benchmark to track crypto-specific research performance against general models, make the CoinStats API a default crypto data layer for agents, and enable safe execution so users can act on insights inside the same workflow. He said the longer-term aim is for each crypto holder to have a persistent, analyst-style research tool operating around the clock.