Buildix MCP Server: Free and API Modes for Live Hyperliquid Orderflow in Claude, ChatGPT and Cursor
The Buildix MCP server runs in two read only modes. Free mode gives any AI assistant five orderflow tools with no account and no key. API mode gives bots and scripts seven plan-gated tools with raw values, CVD history, HIP-4 edge and the liquidation map.
$ Stop reading delayed data. Read live order book depth on the 100 most liquid Hyperliquid pairs right now.
Launch Free Terminal →The Buildix MCP server connects AI assistants and agents to live Hyperliquid orderflow analytics through the Model Context Protocol. It comes in two modes. Free mode runs at https://www.buildix.trade/api/mcp/public with five read only tools and needs no account and no key. API mode runs at https://www.buildix.trade/api/mcp with seven read only tools, uses your Buildix API key and returns the same raw values as the REST API. Neither mode can place an order or move funds.
What the Model Context Protocol is
The Model Context Protocol, or MCP, is an open standard for connecting AI models to outside tools and data. An MCP server publishes a set of tools, and any MCP client can discover and call them during a conversation. Instead of pasting numbers into a chat by hand, the model calls a tool and gets structured data back, with the time of the data it was computed from.
Free mode: five tools, no key
Free mode is built for people who ask questions in a chat. Every answer is a short reading the assistant can quote, and every tool accepts detail: "brief" (the default) returns labels and the key numbers in a few hundred tokens, "full" adds the underlying values, the age of each source and a short note on method. Each answer ends with the line "Informational orderflow analytics, not financial advice."
| Tool | Input | What it returns |
|---|---|---|
| get_orderflow_snapshot | symbol, detail | Direction of recent buy and sell flow (CVD), order book imbalance (OBI) and flow toxicity (VPIN) as categories, plus funding and open interest for one market |
| get_liquidation_levels | symbol, detail | The main liquidation levels above and below the price, estimated from the whole open interest; in full detail a wider range and the amount liquidated on moves of 5, 10 and 15 percent |
| get_whale_positioning | symbol, detail | Aggregated long and short notional of the large Hyperliquid accounts Buildix tracks, the number of accounts on each side and the resulting bias, with no addresses |
| get_funding_rates | detail | Funding ranked from highest to most negative, annualized, with a market wide summary; markets under $1M of open interest are left out |
| get_market_screener | detail, sort_by | Number of markets, total open interest and 24h volume, breadth, open interest weighted funding, top markets and the largest 24h gainers and losers |
Symbols follow Hyperliquid naming: BTC, kPEPE, and HIP-3 builder markets such as xyz:NVDA.
API mode: seven tools, plan-gated
API mode is built for bots and scripts. Each tool sits at the same plan as its REST equivalent and spends from the same daily quota, so a plan includes the same data whichever protocol you use. API keys are issued on the Whale and API plans.
| Tool | Plan | What it returns |
|---|---|---|
| get_screener | no key needed | Every tradable Hyperliquid perpetual with mark price, 24h change, 24h volume, open interest in USD and funding, raw and annualized |
| get_pair | no key needed | One pair in detail: mark, oracle, 24h change, volume, open interest, funding and a deepViewUrl into the Buildix orderflow workspace |
| get_signals | Whale, API | Current signals from the Buildix V5 and V6 orderflow engines across all pairs: direction, signed score, confidence 0 to 100, regime, gated components and up to six reasons |
| get_smart_money | Whale, API | Tracked whale wallets ranked by account value, their largest open positions and the long or short bias of the coins most of them hold, from a snapshot refreshed every 10 minutes |
| get_liquidation_map | Whale, API | Liquidation clusters with the estimated size in USD at each price level around the mark, plus cascade scenarios for moves of N percent; HIP-3 markets as xyz:BRENTOIL |
| get_hip4_edge | Whale, API | Active HIP-4 outcome markets ranked by the gap between model and market: implied and model probability, divergence, an edge_score blended with the V5 signal on the underlying perp, and an edge_flag |
| get_cvd | Whale, API | Cumulative volume delta as a time series at 5m, 15m or 1h, with buy_usd, sell_usd, delta, running cvd and trade count per bucket |
The CVD series is computed on a sample of the tape, a rolling buffer of roughly 300 fills per symbol, so read it as a directional measure and as the basis for price versus delta divergence, not as total traded volume. The payload says so in sampling_note and reports the real coverage_minutes.
Send the key on every request, in either header:
` Authorization: Bearer bx_your_key_here x-api-key: bx_your_key_here `
Assistants whose connector form takes only a URL, such as Claude, ChatGPT, Le Chat, Perplexity and Grok, accept the key on the URL instead: https://www.buildix.trade/api/mcp?api_key=bx_your_key_here. Headers are read first. A key in a URL can end up in client logs and browser history, so give that connector its own key and revoke it from the dashboard if it is ever exposed.
A request without a key resolves as anonymous and reaches only get_screener and get_pair. A tool your plan does not cover answers inside the MCP response with a readable message that names the tool and the plan it needs, so the assistant can explain it instead of failing silently. The daily quota resets at Europe/Rome midnight: 2,000 requests on Whale and 50,000 on API, shared with the REST API.
Which one to use
Use free mode when you ask questions in a chat: pressure, liquidation levels, whale bias, funding and a market overview, with nothing to configure beyond one URL.
Use API mode when you build a bot or a script: raw values in the same shape as the REST API, the CVD history as a time series, the signal engine, the HIP-4 edge ranking and the liquidation map with cascade scenarios.
How to connect
Both modes speak Streamable HTTP over POST and are stateless, so there is nothing to install. In Claude Code:
` claude mcp add --transport http buildix https://www.buildix.trade/api/mcp/public claude mcp add --transport http buildix-api https://www.buildix.trade/api/mcp --header "Authorization: Bearer bx_your_key_here" `
In Cursor, add an entry to ~/.cursor/mcp.json:
`json { "mcpServers": { "buildix": { "url": "https://www.buildix.trade/api/mcp/public" } } } `
In Claude, ChatGPT, Le Chat, Perplexity and Grok, add a custom connector and paste the URL of the mode you want; neither mode uses OAuth. VS Code takes a .vscode/mcp.json entry with "type": "http". The current steps and snippets for every client, in both modes, are on the integrations page at https://www.buildix.trade/integrations#api.
Why an MCP server for orderflow
Most crypto data that reaches an AI agent is stale, scraped or limited to spot prices. The Buildix MCP server hands the model the same orderflow layer that powers the screener: cumulative volume delta, VPIN, order book imbalance, whale positioning, liquidation structure and the signal engine that scores each pair from minus 100 to plus 100. Answers come from data Buildix has already collected, so a call returns in well under a second and never fills a gap with an older figure.
This pairs with the Buildix AI Strategy Advisor, which supports your own keys across several model providers. The advisor brings your model into Buildix; the MCP server brings Buildix into your model.
Frequently asked questions
Is the Buildix MCP server free? Free mode is: it needs no account and no key and has a shared daily capacity. API mode uses a Buildix API key, issued on the Whale and API plans.
What is the difference between the two modes? Free mode returns short readings for a chat. API mode returns raw values for code, plus tools free mode does not have: signals, smart money wallets, the liquidation map with cascade scenarios, HIP-4 edge and the CVD time series.
Which AI clients can connect? Claude, Claude Code, ChatGPT, Le Chat, Perplexity, Grok, Cursor, VS Code and any other client that speaks MCP over Streamable HTTP.
Can it place trades? No. All tools in both modes are read only by design.
How do I pass my API key? In the Authorization: Bearer or x-api-key header, or as ?api_key= on the URL when the client offers no header field.
Where are the full docs? At https://www.buildix.trade/integrations, with copy and paste setup for every client in both modes.