Dataline
Login
live

Supercharge your AI agent
with clean Web3 data

Across Hyperliquid, Polymarket, every major DEX, and the Web2 sources agents need

Join waitlist →See how it works
dataline / 026
04 · 29 · 2026
19.4M
on-chain transactions
96.4%
agent task success rate
334.8K
global community
3
live chains · BNB · Sui · TON
Trading Markets
Hyperliquid
L2 orderbook · perps · vault
+ also reading
BinanceCoinbasedYdXUniswap
Prediction Markets
Polymarket
markets · odds · resolution
+ also reading
KalshiManifold

Built with, backed by

/ built with
LangChainFlowAgent backbone
Cetus ProtocolSui DEX integration
Numbers Protocolverifiable outputs
RootDataWeb3 dataset
/ backed by
Googlefor StartupsCloud Program
BNB Chainecosystem partner
BNB AI Hack2025 winner
/ [01] — product stack

Start with the data API
Add the rest when you need it

Read every market with Dataline Data — orchestrate multi-step agents with FlowAgent, click through long-tail dApps with GhostDriver, deploy to Telegram with ChatPilot

{}
Flagship · Data + execution API

Dataline Data

Clean, confidence-scored data from every crypto market — with a non-custodial Act layer on the same surface. Read with confidence. Act on every venue.

  • Read every market with confidence scoring
  • Act non-custodially on any venue
  • MCP-native + REST + SDKs
Explore Dataline Data →
/ also in the stack
Orchestration runtime
FlowAgent

Orchestration runtime for multi-step agent workflows on hosted LangGraph + Dataline primitives.

For: devs building multi-step agents
Browser-exec fallback
GhostDriver

Browser-execution fallback for long-tail dApp UIs when no API exists.

For: devs needing long-tail UI coverage
Consumer chat UI
ChatPilot

Consumer chat UI — deploy a Telegram chatbot for crypto in one click.

For: community managers + end users
/ [02] — pipeline

Intent in
Confidence-scored data out

Five stages, one API call — each stage is testable, observable, replaceable

/ 01

Intent

Describe what you need in plain language

“BTC price”
“ETH funding”
“election odds”
/ 02

Route

We pick the best sources automatically

smart routing
multi-provider fallback
latency optimized
/ 03

Normalize

All data cast to one schema

no vendor formats
same shape across
every venue
/ 04

Aggregate

Reliable production output

outlier filtering
median / blended
derived metrics
/ 05

Output

Clean JSON with confidence score

{ symbol, price,
▍, ts }
/ [03] — design principles

Built for the way agents actually consume data

+ unified

One unified API

Every crypto data source under one schema. Switch venues without rewriting the agent.

+ confidence

Confidence on every call

Routing, fallback, normalization — with a score so your agent can refuse unreliable data before acting.

+ act

Read AND act

Non-custodial execution included — swap, perp, bet, vote. Your agent signs; your wallet broadcasts.

+ mcp

MCP-native

Plug into Claude, Cursor, Windsurf, eliza in 30 seconds. JSON-Schema-clean tools.

+ pricing

Pay how you want

Subscription, X402 (Basechain pay-per-call), MPP (Tempo) micropayments — three rails, your choice.

+ tested

Production-tested

19.4M tx executed at 96.4% success across BNB, Sui, TON.

/ [04] — use cases

From idea to production agent
in a weekend

trading

Polymarket ↔ Hyperliquid arb

Same event, two venues, auto-execute the spread.

trading

Hyperliquid vault copy-trader

Sub-second mirror of any top vault, with risk caps.

signal

News-to-trade

Headline drops → PM or HL perp moves first → take the lag.

signal

Memecoin radar

pump.fun graduations × smart-money inflows, alert in <5s.

risk

Liquidation hunter

Watch HL + Aave health, snipe at the wick.

research

DAO governance agent

Read every Snapshot proposal touching X, with treasury exposure.

npm install · pip install · MCP add-server

npm
pip
curl
mcp
/ [05] — FAQ

Questions, answered

Why not just call Coingecko and the Hyperliquid SDK directly?
You can — for one source. Dataline aggregates across many sources, normalizes them into one schema, and ships every response with a confidence score so your agent can detect divergence and refuse stale data. Building that yourself is a multi-week integration per source.
What chains and venues do you cover at launch?
Hyperliquid and Polymarket are the launch flagships. Underneath: BNB Chain, Sui, TON (where 19M+ tx already run), plus Coinbase, Binance, Kalshi for market data. More chains roll out monthly.
How does MCP work with Claude / Cursor / Windsurf?
One add-server command in your host. Your agent reads our published AGENT_SKILL.md, registers tools, and starts calling. Every tool ships with JSON Schema, source attribution, and per-call pricing visible in the response.
How does the Act layer stay non-custodial?
Dataline returns a signed transaction; your wallet broadcasts. Funds never leave the user's custody. On Hyperliquid we use trade-only API permissions. Your agent acts; you stay in control.
How does Dataline compare to other Web3 data providers like The Graph, Dune, or Alchemy?
Those are query layers over indexed historical state — strong for analytics dashboards, slow for live agent decisions. Dataline is the runtime layer your agent hits when it needs to act now: live aggregation across exchanges, chains, and prediction markets, returned as one schema with a confidence score. We don't replace your warehouse — we sit between your agent and 22+ live sources so it can decide and execute in one round-trip.
What's the latency?
P50 ~87ms for cached intent calls, ~250ms for fresh aggregation across multiple sources. Streaming endpoints (orderbook, ticks) deliver under 30ms. Hyperliquid execution adds ~120ms median for signed-tx return. Routes pre-warm against your top intents so the first call of the day isn't your slowest.
What happens if a source goes down?
The router rebalances in real time. If Binance drops out, weight shifts to OKX, Coinbase, and Kalshi automatically — your call still returns. The confidence score reflects the reduced consensus so your agent sees the degradation in the response and can refuse, retry, or proceed with caution. Single-source responses always carry a flag.
Is there a free tier? Do I need a credit card?
Yes. 500 free calls, no credit card required to start — you can hit the resolve endpoint in under 30 seconds. Hobby tier starts at $19/mo with 50K calls and MCP access. Pro and Enterprise add execution, trace logs, and dedicated routing. See /pricing for the full breakdown.

Stop stitching APIs
Start shipping agents

Free tier, no credit card, 500 calls — see the response shape in 30 seconds

Read Docs →