Agents need one reliable data plane, not a pile of adapters.
Market, on-chain, prediction-market, and news data arrive through incompatible APIs and schemas. Teams repeatedly rebuild routing, normalization, quality checks, and attribution. The protocol’s premise is that those capabilities should be shared infrastructure.
| Need | Current gap | Protocol response |
|---|---|---|
| Price and market data | Multiple APIs and inconsistent schemas | One normalized request and response model |
| Perpetuals and DeFi | Siloed SDKs and venue-specific fields | Intent-based routing across supported venues |
| Prediction markets | No shared identity or aggregation layer | Cross-venue results with source-level freshness |
| Quality | Custom ETL and hidden disagreement | Traceable sources, coverage, and confidence metadata |
From intent to evidence-backed output.
- Intent ParseResolve the market, time range, and caller constraints.
- RouteSelect the appropriate tools and eligible sources.
- NormalizeAlign identifiers, units, decimals, timestamps, and fields.
- AggregateCompare sources, filter anomalies, and retain provenance.
- OutputReturn API-, MCP-, or interface-ready data with quality metadata.
Confidence is a versioned score informed by agreement, freshness, coverage, and source reliability. It is not a fixed multiplication formula and does not guarantee correctness.
Supply, verification, and incentives form one quality loop.
Providers contribute data and stake against service quality; demand comes from agents and applications. Validation, reputation, optional zero-knowledge proofs, rewards, and slashing are intended to align availability and correctness over time.
A one-billion-token allocation.
| Allocation | Share | Tokens |
|---|---|---|
| Community rewards | 33% | 330,000,000 |
| Ecosystem and foundation | 30% | 300,000,000 |
| Launchpad and liquidity | 13% | 130,000,000 |
| Team | 12% | 120,000,000 |
| Investors | 12% | 120,000,000 |
| Total | 100% | 1,000,000,000 |
Whitepaper projection: 21.5% unlocked at token generation and 100% by month 48. These are project-design figures, not market or investment guidance.
Four layers, with explicit operating boundaries.
Dataline Data
- Responsibility
- Shared, chain-agnostic data foundation that routes intent, normalizes multi-source results, and returns confidence metadata.
- Use cases
- Research, agents, bots, dashboards, read-only wallet queries, and unsigned transaction preparation.
- Boundary
- No custody, private-key handling, or transaction signing.
FlowAgent
- Responsibility
- Orchestration and control plane for planning, routing tools, evaluating confidence, and composing results.
- Use cases
- Multi-step DeFi workflows, trading agents, advanced research, and embedded product flows.
- Boundary
- No unbounded loops; risky, ambiguous, or low-confidence fields pause for clarification or approval.
GhostDriver
- Responsibility
- Browser-native execution layer for workflows without a reliable API, operating in the user’s real session.
- Use cases
- Airdrops, DeFi sites, wallet cleanup, and cross-platform browser workflows.
- Boundary
- Signatures, transfers, and irreversible actions remain under explicit user control.
ChatPilot
- Responsibility
- Conversational product surface for Ask, Swap, dashboards, and guided workflows.
- Use cases
- Market questions, portfolio views, deterministic task routing, and access through web or Telegram.
- Boundary
- Non-custodial; the user approves wallet signatures and irreversible operations.
Readable alternatives to the architecture diagrams.
Dataline Data
Request → intent router → source executors → normalization → aggregation and confidence → cache → REST / MCP output
Chain-agnostic identifiers and per-source metadata keep results traceable. Read-only queries and unsigned transaction construction preserve the non-custodial boundary.
FlowAgent
Intent → Router → Plan → Tool graph → Decision and confidence gate → Answer / UI
The graph makes intermediate decisions inspectable and pauses when inputs, confidence, or approval are insufficient.
GhostDriver
Task → page recognition → policy and sitemap → browser action plan → user approval gate → execution evidence
Execution happens in the user’s browser session; sensitive and irreversible actions remain user controlled.
ChatPilot
Conversation → deterministic route → Dataline Data / FlowAgent → grounded response or guided action
Memory and interface rendering improve continuity without moving wallet-signing authority away from the user.
MCP service families
| Family | Purpose | Selection boundary |
|---|---|---|
| Market data | Prices, candles, derivatives, and funding | Use for quantitative market questions; retain venue and time context |
| DeFi protocol data | Pools, liquidity, yields, and protocol state | Use for supported protocols; do not infer missing chain or token metadata |
| Information aggregation | News, prediction markets, and cross-source research | Use when synthesis and provenance matter; surface disagreement |
| Transaction execution | Prepare supported transaction workflows | Return unsigned or approval-gated actions; never claim custody |
| GhostDriver | Browser workflows without a reliable API | Use only inside the user-controlled session and approval policy |
Performance and data targets
| Area | Whitepaper target | Interpretation |
|---|---|---|
| Latency | Median <87 ms; P95 <150 ms; P99 <300 ms | Milliseconds measured at the service boundary |
| Capacity | 10,000 sustained / 50,000 peak requests per second | Engineering target, not a per-customer guarantee |
| Freshness | Price <5 s; liquidity <15 s | Seconds from source observation; historical results may be precomputed |
| Reliability | 99.95% uptime; <0.1% error rate | Confirm the production measurement window |
| Accuracy | >99.5% when validated | Only where a validation benchmark and source coverage exist |
Figures reported in the whitepaper.
This is a whitepaper snapshot, not live telemetry. Source dashboards, metric definitions, and measurement windows should be reviewed during diligence.
Foundation → launch → expansion → scale → future.
- Foundation / LiveCore protocol and product foundations
- Launch / NowPublic integrations and distribution
- Expansion / Near-termBroader data, tools, and partner coverage
- Scale / UpcomingCapacity, reliability, and network growth
- Future / AheadOpen-network and ecosystem development