On-chain Data Capabilities
xFractal is the first NLP engine that actually understands Solana's on-chain dynamics.
We're a dynamic transponder. We listen to the solana event horizon in real-time, and reflect meaningful info back to the user. Think of it as a socket-powered, agent-led, streaming NLP for on-chain data. Down to the microcaps.
That’s why the in-dApp alpha extraction methodology hits different; we react in real-time, adapt instantly, and deliver exactly what matters. You get filtered, context-aware intelligence. fast, accurate, actionable.
gRPC channels tap directly into the solana event flow. xFractal listens, parses, and routes data through attention-based sockets. Historical data is fine. But nothing competes with real-time dynamic agents actively watching and understanding the chain.
The problem?
Solana emits data like a black hole. It’s constant, directionless, and unfiltered. Each emission is a raw byte array. Unreadable, noisy, and impossible to decode without proper parsing.
Our solution?
Buckets. Buckets are localized partitions of the data stream. They chop the waterfall of emissions into usable, interpretable chunks. Each bucket is scoped to a specific context:
E.g., Swaps, token launches, top holders, dev wallets, bundlers, LP inflows, etc.
They’re dynamic, time-bound, and self-cleaning.
Some hold 5 mins of emissions, others 15. Older data gets flushed. New data flows in. You can listen to buckets in real-time or request historical payloads.
Buckets are composable. An agent can cross buckets to follow relations:
Token → top holders → bundlers → LPs launchpad → token created → wallet → bundled supply
You can filter buckets. Chain them. Route outputs to other buckets. Even redirect a query if the data you want isn’t here, but two hops away. The user never sees this, but the NLP and agents do. they make the request, follow the chain, and deliver a tight, clean objective answer + their subjective opinion based on the data they have access to, their historical context, and ecosystem training.
Fully modular. Fully real-time. Agents like Echo + Mobu talk to each other via handles and pass data to one another. One request → multi-layered response.
A different data pipeline
We listen to everything. Here's how we achieve this:
Attention-based listeners: xFractal keeps listening to active programs and accounts.
Modular receivers: ingest a broad spectrum of data to filter, parse, and sort.
Waterfall mechanics: progressive parsing to present relevant and meaningful data.
Core components:
Bucket Services: Create data partitions for programs, events, and accounts. A service handler fetches the most relevant data in the operation.
Single bucket integration: lump sum events for changes over time: (1m, 5m, 1h, 4h, 12h, 24h), price, MC, FDV, volume, makers - buyers + sellers, transactions - buy + sell, volume
Partial / cross bucket operation: count # of accounts using a relation (e.g. token → # top holders → # bundlers → LP )
External monitoring: P&L performance - realized + unrealized (24h, 7d, 30d, 1y), token bundling (e.g. # bundles, % Supply bundled, bundler wallet analysis), and hot programs: list of programs with top activity for an event (e.g. swaps, token creation, etc)
Internal monitoring: monitor usage, downtime, and KPIs.
Data Standards:
Freshness: data is instant and on-demand
Integrity: data is filtered with minimal overhead
Relevance: data is parsed for meaningful results
On-chain data dashboards can be found within Mobu (on-chain analyst) and Sentra (security scanner) agents.
Mobu (On-chain Analyst):
Sentra (Security Scanner):
Last updated