Surf AI

Surf AI (styled Surf, and operating at asksurf.ai) is a cryptocurrency-focused artificial intelligence platform developed by Cybertino Inc. that packages market data, on-chain analytics, prediction-market feeds, company financials and news into conversational research tools and developer APIs. The company describes Surf as "Crypto's Ultimate AI" and a "Trusted Intelligence Layer for Digital Asset Institutions," positioning the product as an AI analyst for open financial markets that returns evidence-backed, sourced answers rather than unattributed text.^[1][2] The corporate entity behind the platform is identified in Surf's own materials as Cybertino Inc., with copyright notices dated 2026.^[3]

Surf states that it serves more than 500,000 users across over 700 organizations, that it has processed more than 6 billion on-chain events, and that it covers over 17,000 crypto projects and more than 75 protocols.^[1] The platform is built around a single shared data layer that feeds a family of consumer products and a set of developer interfaces, all served by Surf's proprietary surf-2.0 model family.^[4][5]

Platform and Architecture

Surf's current generation is branded Surf 2.0, which the company launched in early access as a platform that combines multi-model orchestration, crypto-native tooling and sector-specific data to complete crypto workflows end to end.^[3] Rather than relying on a single large language model, Surf states that it runs on more than ten AI models, drawing on systems from xAI, OpenAI, Anthropic, Google, DeepSeek and Kimi alongside its own custom fine-tuned models. The company says orchestration between these models is automatic and that they are extended with proprietary tooling to handle crypto-specific tasks such as querying on-chain records and analyzing social sentiment. Surf frames this as its core differentiation, stating "It's not one model, it's several. All orchestrated with specialized data, applying domain-informed reasoning, and using custom tools."^[3]

Surf 2.0 is exposed through two primary surfaces. Surf Studio is a no-code builder aimed at people, letting users turn natural-language prompts into custom crypto web apps without configuration or hosting. Surf Agent Stack (SAS) is the developer-facing surface, described as a set of APIs, Skills and an MCP server that opens Surf's data and reasoning as infrastructure for AI agents, packaged for integration into environments such as Claude Code and OpenClaw. According to the company, SAS added more than 60 new API endpoints across seven categories.^[3]

Data Layer

All of Surf's products draw from one common data layer, which the company organizes into distinct domains and describes as spanning 12 data domains and more than 23,000 tracked projects.^[4] The company states the platform draws from more than 30 integrated industry-specific data sources together with its own proprietary database, and that its research engine connects more than 200 sources across 40 or more chains as well as US equities.^[3][2]

The market-data component provides real-time prices sourced from 16 exchanges, pre-computed technical indicators including RSI, MACD, Bollinger Bands, NUPL, SOPR and MVRV, ETF flows for spot Bitcoin and Ethereum products, and options data with Greeks, spread across 18 endpoints.^[4] An on-chain SQL layer is described as offering more than 80 analyst-ready tables with sub-second queries covering DEX trades across seven chains, token transfers, total value locked (TVL), lending flows, staking and bridges, with agents able to auto-discover table schemas so queries do not need hardcoding.^[4] A Wallet Profiler returns multi-chain balances, DeFi positions across protocols such as Aave, Lido, Uniswap, Compound, Curve and Morpho, and transaction history with token-level detail.^[4]

Surf's address-intelligence layer is described as covering more than 100 million labeled addresses across 13 networks, with entity-specific labels — the company cites examples such as "Binance Hot Wallet 8," "Wintermute Trading Desk" and "Jump Trading Address 3" — rather than generic categories.^[4] Its prediction-market layer covers both Polymarket and Kalshi with cross-market matching, and the company states it holds more than 934 million rows of trade data along with order-book depth and whale position tracking.^[4] A funding and venture-capital domain covers rounds from seed through Series C with investor lists, valuations, lead investors, fund portfolios and rankings, while an AI news domain aggregates more than 30 sources — including The Block, CoinDesk, Cointelegraph, Decrypt and Blockworks — with AI classification by sentiment and relevance and original briefs branded "Surf AI News."^[4] Elsewhere the company reports coverage metrics including more than 29.4 billion token transfer records across chains, over 120 million social posts indexed with more than 200,000 accounts monitored in real time, over 16,000 crypto projects tracked in Cryptopedia and more than 12,000 whale and key wallets monitored in real time.^[3]

Products

Surf markets a group of consumer products — the company describes the lineup as five products running on the same data and the same intelligence — that share the underlying data layer while addressing different tasks: answering questions, building apps, looking up projects and following the market.^[4]

Surf Chat and Research

Surf Chat is the core conversational agent, described as pulling from all 12 data domains simultaneously to answer crypto questions with live, cross-referenced and sourced data. Example queries include retrieving the funding history of EigenLayer, checking Ethereum whale activity over the previous 24 hours, or comparing Arbitrum and Base DEX volume over 30 days.^[4] The research experience is delivered through a product Surf calls Research, which plans the work, pulls market, on-chain, filings and news evidence, cross-checks its findings and returns a single sourced report drawing on more than 200 connected sources across 40 or more chains plus US equities.^[2]

Research runs in two modes. Instant Research is built for fast facts and market checks, returning prices, metrics, definitions and quick comparisons with inline citations in what the company says is ten seconds or less. Deep Research runs a structured investigation with a visible plan and live progress, offered at four depth levels — Standard, Extended, Maximum and Ultra — and delivering a decision-ready report with findings, risks, counterarguments and citations.^[2] Surf exposes across these modes market data such as live price, spot and perpetual volume, open interest, funding and liquidations; on-chain data including wallet balances, transfers, labels, DeFi positions and token flows; equities data covering US stocks and ETFs, financial statements, ratios and SEC filings; fundamentals such as teams, investors, funding, tokenomics and unlock schedules; and news, governance and prediction-market odds.^[2]

Surf Studio

Surf Studio is a no-code builder that turns a natural-language description into a working crypto web app connected to Surf's live data layer, allowing users to preview, tweak and deploy to a live URL. The company gives example prompts such as "build me a whale alert dashboard" or "create a Polymarket vs Kalshi arb scanner," and states that every generated app ships with real-time crypto data — prices, wallets, on-chain and prediction markets — already wired in, claiming a path from "zero to production in 5-10 minutes."^[4] Live examples of apps built with Surf that the company has cited include a prediction-market geopolitics dashboard, a crypto wallet forensics app, a "VC Report Card" and a Bittensor ecosystem map.^[3]

Cryptopedia and Crypto Pulse

Cryptopedia is Surf's structured library of crypto projects, which the company describes as covering more than 23,000 projects, each broken down by on-chain activity, funding history, tokenomics, team, DeFi metrics and airdrop status and cross-referenced against live data.^[4] Cryptopedia incorporates a Signal Map that scores every price move against an asset's own history so that unusual moves rank above ordinary ones, ranking the market's largest anomalies and naming the driver of each. It draws on ten live signal inputs — including price, spot and perpetual volume, open interest, funding and liquidations — and organizes them into four intelligence dimensions labelled Movement, Flow, Derivatives and Attention, producing explainable output that shows what changed, why it is unusual and which inputs drove it, along with a signal direction indicating whether an anomaly is strengthening or fading.^[2] Selecting a signal opens the full project record, described as five intelligence layers with more than 50 data points per project spanning live markets and derivatives, team, funding, tokenomics, news and prediction-market context; for US-listed names the record adds company profile, financial statements, ratios and SEC filings.^[2]

Crypto Pulse is a live news product that pairs a daily event feed of more than 1,000 items with more than 80 AI-written briefings per day, and supports six languages. The company states the real-time feed catches funding rounds, listings, releases, price events and influential posts as they land, and that the news is AI-classified by sentiment — bullish, bearish or neutral — and relevance, with no algorithmic timeline, engagement bait or promoted content.^[2][4]

Developer Infrastructure

For builders, Surf exposes its data and reasoning through the Surf Agent Stack, comprising a Chat API, a Data API, a Skill for AI coding agents, an MCP server and a command-line interface. The company summarizes three ways to access the platform: a one-line install for AI coding agents; a Skill offering more than 100 commands that takes natural language in and returns structured data out; and roughly 120 public RESTful endpoints across 17 domains for direct programmatic access, including nine equity endpoints — all unified under a single API key covering more than 40 blockchains, over 200 data sources and US-listed equities.^[5]

Chat API

The Chat API embeds the same conversational crypto research experience found at asksurf.ai into external applications, returning fully sourced natural-language answers that can include live prices, on-chain activity, social signal, ETF flows and macro context.^[6] It is served by a component called Research 2.0 at the OpenAI-compatible endpoint POST /v1/responses, powered by the surf-2.0 model family and fronted by a proxy named Hermod that handles authentication, credit billing and request routing.^[7] The company markets Research 2.0 with claims of "faster answers, deeper coverage, more reliable reasoning," including four times fewer hallucinations, twice the answer depth and twice the research-logic consistency relative to the prior generation.^[7]

Two model variants are offered. surf-2.0-instant is the faster, lower-latency option and always costs 20 credits per call, while surf-2.0 is described as the full research engine whose cost scales with a reasoning.effort parameter: an omitted or "none" effort costs 20 credits, "minimal" 30, "low" 50, "medium" 120, "high" 150 and "xhigh" 200 credits per call. The documentation recommends "medium" effort for most questions, "low" or "minimal" for latency-sensitive lookups, and "high" or "xhigh" for complex open-ended research.^[6][7] The prior-generation Research 1.5 endpoint, POST /v1/chat/completions, has been retired and now returns HTTP 410 Gone; clients are directed to migrate by moving their prompt into the input field and nesting reasoning depth under reasoning.effort, since the flat reasoning_effort field is no longer accepted.^[6][7]

Data API

The Data API provides typed REST endpoints for crypto market data, project analytics, on-chain queries, US equities, company financials and ETF holdings, each billed per request in credits and rate-limited per API key.^[8] Endpoints are grouped by domain, including exchange data (last traded price, OHLCV candlesticks, live order-book depth, and perpetual-specific funding history and long/short ratios); market endpoints (CoinGecko-sourced hourly prices, rankings, BTC-centric on-chain indicators such as NUPL, SOPR and MVRV, and daily flow data for US spot BTC and ETH ETFs); token endpoints for DEX trades, transfers and top holders; wallet endpoints for balances, DeFi positions, labels and historical net worth; project endpoints for DeFi metrics and rankings; fund endpoints returning fund profiles, portfolios and rankings; prediction-market endpoints for Polymarket and Kalshi; news, search and web-fetch endpoints; and on-chain endpoints that allow single transaction lookups, structured queries, and arbitrary SQL against 106 ClickHouse tables across eight chains.^[8] Supported chains identified by canonical name include Ethereum, Polygon, BNB Chain, Solana, Avalanche, Arbitrum, Optimism, Fantom, Base, Linea and Cyber.^[8]

A distinctive feature is opt-in address-label enrichment: six endpoints accept an include=labels parameter that attaches entity-label records — covering exchanges, protocols, funds, market makers, notable wallets and ENS-registered addresses — to every address in a response. The feature is additive, so responses without the parameter remain byte-identical to the pre-feature format; labels are cached per address for roughly one hour, and a 200-millisecond timeout wraps each enrichment call so that a slow label backend returns unlabeled items rather than failing the request.^[8]

Skill, CLI and Equity Endpoints

The Surf Skill is installed into AI coding agents such as Claude Code and Cursor with a one-line command (npx skills add asksurf-ai/surf-skills --skill surf), giving an agent more than 100 crypto data commands that it can call in natural language.^[5] The documented Skill domains and command counts include Prediction Market (23 commands for Polymarket and Kalshi events, trades, prices and analytics), Market (15 commands), Search (12), Exchange (9), Hyperliquid (9 for trader positions, fills, orders and leaderboards), Onchain SQL (8 commands over 106 ClickHouse tables across eight chains), Wallet and Token (6 each), Project (5), Signal (4 for token heat scores), Fund (3), News (2), and single-command DEX and Web domains.^[5] The company states the Skill "teaches Claude Code and Cursor to query Surf's CLI natively — no hardcoding."^[4] The Data API also adds nine public equity endpoints for US stocks and ETFs, split between market-context endpoints (price candles, listings and ETF holdings) and company-fundamentals endpoints (profiles, overview, income statement, balance sheet, cash flow and metrics).^[5]

Access, Credits and Pricing

Surf uses a credit-based billing model applied across its APIs. Every IP address receives 30 free credits per day without signing up, resetting at 00:00 UTC, after which unauthenticated requests from that IP are blocked until the next reset. Full access requires signing in with Google or email at agents.asksurf.ai, generating an API key from the Dashboard — keys begin with the prefix sk- — and topping up by credit card or with USDC.^[8] Most Data API endpoints are priced as Light (1 credit), Standard (2 credits) or Heavy (4 credits) calls, with equity financial statements costing 6 credits; re-calling the same endpoint within three minutes is free, and every response reports its cost under meta.credits_used.^[8]

Rate limits are enforced per API key over a sliding 60-second window, with each account holding a single bucket regardless of how many keys it issues. The default pay-as-you-go tier, assigned automatically on sign-up, permits 100 requests per minute per key and allows bursts, while an enterprise tier is configured per contract based on expected throughput. Exceeding the limit returns HTTP 429 with the error code RATE_LIMITED. The documentation specifies a full set of error codes and HTTP mappings, including FREE_QUOTA_EXHAUSTED and PAID_BALANCE_ZERO for credit conditions, upstream error codes for temporarily unavailable data sources, and QUERY_TIMEOUT when a ClickHouse SQL query exceeds its time limit.^[8]

Consumer access is organized into subscription tiers. The company lists Pro, Plus and Max plans; Surf Chat is available on the Pro and Max tiers, with Pro priced at $29 per month, while Surf Studio is a Max-tier exclusive with Max pricing shown as a range of $99 to $1,999 per month.^[4] At launch, Surf Studio was made available to Max subscribers with Pro and Plus users receiving a limited preview, and Max users were given the option to try the Agent Stack by claiming an API key with trial credits.^[3] Surf's FAQ addresses common subscription and product questions, including image upload support, supported languages, cancellation and whether Surf has a token.^[9]

Adoption and Clients

Surf reports use across trading firms, exchanges, research teams and venture funds, and states that "teams worldwide run on Surf."^[1] Among its cited engagements, the company says it worked with a local exchange team to build an AI intelligence layer into the exchange's end product that analyzes a million traders' portfolios daily, delivered bespoke crypto data and research services to an investment research team, integrated its Chat API into Defined's Codex to give users DEX intelligence within seconds, extended the investment firm IOSG's internal AI research agent with five new categories of bespoke crypto data, and unified fragmented multilingual Asian centralized-exchange data into a single structured dataset for a research team.^[1]

Surf publishes testimonials from named industry figures, including Heechang Kang, chief strategy officer and co-founder of Four Pillars, who said Surf "turns raw crypto news into digestible research at a speed no traditional workflow can match"; Danning Sui of Pantera Capital's research team, who described Surf's MCP and Skills as making "blockchain data seamless in my daily AI workflow"; and Yiping Lu, AI lead at IOSG, who said Surf "delivers datasets the others can't provide."^[1] The company frames its development approach around what it calls closed-loop evaluation, stating that its evaluation process "continues until the client is happy with the output, every single time," and asserts that its security is built for institutional workflows to protect the data, systems and workflows firms rely on when delegating high-stakes research to AI.^[1]