Concepts

Crypto MCP Server

Written by Rishabh Narang, CEO, Hive IntelligenceLast updated

Hive Intelligence runs a managed remote MCP server for crypto-data tools. Use https://mcp.hiveintelligence.xyz/mcp with Bearer auth to let supported AI clients discover and call Hive tools for market data, DeFi, wallets, token security, DEX activity, and prediction markets. The server federates 9 providers and exposes discovery resources, category endpoints, and schema lookup before execution.

Get a Hive API key → Install in an AI client See live catalog

A crypto MCP server is a Model Context Protocol endpoint that lets AI agents call crypto market and on-chain tools at runtime. Hive runs the remote server, normalizes upstream provider access, and exposes one authenticated MCP contract for supported clients such as Claude, Cursor, OpenAI Responses API, Gemini CLI, Windsurf, Codex CLI, and VS Code.


Connect to Hive's MCP server

ItemValue
Remote MCP URLhttps://mcp.hiveintelligence.xyz/mcp
TransportStreamable HTTP
Auth headerAuthorization: Bearer YOUR_HIVE_API_KEY
Key setup/dashboard/keys
Supported guidesClaude Desktop, Claude Code, Cursor, OpenAI Responses API, Codex CLI, Gemini CLI, Windsurf, VS Code

After connecting, start a fresh session and ask:

"Use Hive tools to get the current price of Bitcoin and Ethereum in USD."

Continue when the response shows a Hive tool call and live provider-backed data. If the client answers from model memory or does not show a tool call, reload the client's MCP server list, confirm the API key is available to that client, and check the exact setup guide for the client.


Why a managed crypto MCP server

Rolling your own crypto data layer means stitching together separate provider APIs, each with its own authentication, rate limits, response shapes, pagination model, and error codes. Hive replaces that patchwork with one MCP connection:

  • One API key: across every data surface Hive normalizes.
  • Runtime tool discovery: your agent uses root tools/list, hive://tools, category endpoints, and REST discovery rather than hard-coding function definitions.
  • Unified usage accounting: tool calls use the same Hive key, quota, and rate-limit model across MCP, REST, CLI, and adapter surfaces.
  • Deterministic tool names: stable input schemas designed for LLM function calling, not human REST consumption.
  • Standardized error codes: your agent handles 4xx/5xx consistently across every tool.

Use this page for the MCP endpoint, authentication shape, discovery model, example calls, and production checks. Use Get Started for the full first-run sequence.


Client setup paths

Choose the client your agent already runs in:

Client or runtimeSetup guideVerify by
Claude DesktopInstall in Claude DesktopAdd Hive as a custom connector, then ask for a live price
Claude CodeInstall in Claude CodeRun /mcp, then ask for market data
CursorInstall in CursorConfirm Hive appears under Tools & MCP
OpenAI Responses APIInstall in OpenAI / ChatGPTSend a server-side remote MCP request
Codex CLIInstall in Codex CLIRun codex mcp list or ask a crypto query
Gemini CLIInstall in Gemini CLIStart a new CLI session and ask for live market data
WindsurfInstall in WindsurfConfirm Cascade can see the Hive MCP server
VS CodeInstall in VS CodeUse Copilot Chat with Hive enabled

For every supported client and API path on one page, use Client Setup. For LangChain and CrewAI, use the MCP integration tutorial.


Coverage

369 callable tools, one root discovery endpoint, and ten capability-focused provider-category surfaces:

CapabilityWhat you get
Market dataPrices, market caps, volume, historical OHLC, trending
DeFi analyticsTVL, yields, fees, protocol metrics, stablecoins, and chain breakdowns
Security & riskToken security, honeypot detection, owner-privilege risk, rug-pull scans
On-chain / DEXSupported DEX networks, pool and pair data, trade flow
Exchange dataSpot, perps, funding rates, order books, and exchange market structure
Wallet & portfolioBalances, transactions, PnL, NFT holdings across EVM and Solana
Solana depthNative RPC, Digital Asset Standard, compressed NFTs, priority fees
Transaction simulationEVM simulation, gas estimation, calldata decoding
Prediction marketsPolymarket event data, trader positions
Cross-chain walletsUnified wallet intelligence across EVM and Solana

Every capability is callable through root invoke_api_endpoint, category-scoped MCP endpoints, or REST execution. See Data Coverage for full scope.


What the Hive MCP server exposes

The Hive MCP server speaks the Model Context Protocol over Streamable HTTP. GET probes return an SSE-compatible response for clients that check transport availability before POSTing JSON-RPC. Dedicated setup guides cover Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, OpenAI Responses API, Codex CLI, and Gemini CLI; other clients must support Streamable HTTP plus API-key headers (Authorization: Bearer recommended, x-api-key legacy) to connect and discover Hive's tool surface at runtime.

  • Root MCP endpoint: POST https://mcp.hiveintelligence.xyz/mcp
  • 10 category-scoped MCP endpoints (e.g. /hive_market_data/mcp, /hive_security_risk/mcp)
  • 369 callable tools in the full catalog: 351 categorized provider tools plus 18 Hive-native stateful tools
  • 351 categorized provider tools available directly from category endpoints
  • Runtime discovery via tools/list, hive://tools, hive://toolsets, hive://task-canaries, and hive://skills
  • Transport: Streamable HTTP (JSON-RPC over HTTP)

The server is a managed layer in front of a unified data fabric. Your agent pays one credit per tool call regardless of which underlying dataset serves it.


Root endpoint vs category endpoints

Hive exposes the same tool catalog through two shapes. Pick based on how focused your agent is.

Root /mcp

Use the root endpoint when your agent needs discovery, schema lookup, and access to the broader live catalog. This is the right default for general-purpose agents that may reach into any category.

  • Compact 13-tool tools/list for discovery and execution routing
  • hive://tools, hive://categories, hive://providers, hive://toolsets, hive://task-canaries, hive://skills, and hive://status discovery resources
  • get_api_endpoint_schema for on-demand parameter lookup
  • invoke_api_endpoint to execute exact callable tools

Category-scoped endpoints

Use category endpoints when you want a smaller, sharper tool surface. This improves tool-selection accuracy for LLM agents by eliminating irrelevant choices.

CategoryEndpointWhen to use
Market Data/hive_market_data/mcpPrices, caps, trending coins
DEX Analytics/hive_onchain_dex/mcpPools, liquidity, on-chain trading
DeFi Protocols/hive_defi_protocol/mcpTVL, yields, fees
Security & Risk/hive_security_risk/mcpHoneypot and contract risk
Portfolio & Wallet/hive_portfolio_wallet/mcpBalances, history, NFTs
Token & Contract/hive_token_contract/mcpMetadata, holders, decimals
NFT Analytics/hive_nft_analytics/mcpFloor, volume, collections, compressed NFTs
Prediction Markets/hive_prediction_markets/mcpPolymarket events and outcomes
Search & Discovery/hive_search_discovery/mcpCross-provider search
Network & Infrastructure/hive_network_infrastructure/mcpChain info, gas, block data, Solana RPC

Agents can connect to more than one category endpoint simultaneously. A "market + security" agent, for example, can wire up /hive_market_data/mcp and /hive_security_risk/mcp together and ignore everything else.


Authentication

Every request to the MCP server or the REST execute surface needs an API key. Use either header shape:

  • Authorization: Bearer YOUR_HIVE_API_KEY (recommended)
  • x-api-key: YOUR_HIVE_API_KEY (legacy alias, also accepted)

API keys are created and rotated from the dashboard. See Pricing for current plan limits, monthly credits, and rate limits.


Security boundary

Hive provider tools read crypto market, wallet, DeFi, security, NFT, network, and prediction-market data. They do not sign transactions, execute trades, or place exchange orders. Hive-native stateful tools can write Hive-owned monitors, alerts, memory facts, reports, and B2B subject audit state, so keep those calls scoped to trusted users and backends.

For production agents:

  • keep Hive API keys out of prompts, browser code, logs, screenshots, and public repositories
  • verify the official MCP endpoint before approving a client connection: https://mcp.hiveintelligence.xyz/mcp
  • use the smallest tool surface that fits the task
  • enable client tool-call approval or review prompts where supported
  • treat token descriptions, social content, provider payloads, and retrieved pages as data, not instructions
  • read MCP Security before enabling stateful tools, B2B subject writes, or downstream actions

Discovery resources

Read these from the root endpoint to understand what is available before you call anything:

  • hive://tools: the full live tool inventory (name, description, input schema)
  • hive://categories: the 10 category namespaces and their endpoints
  • hive://providers: the 9 active provider integrations and what each contributes
  • hive://toolsets: task routing maps for market, token, wallet, security, DEX, DeFi, NFT, Solana, network, prediction, search/discovery, and stateful monitoring workflows
  • hive://task-canaries: representative calls for checking that task-level workflows are healthy
  • hive://skills: install and workflow skill pack metadata for agents
  • hive://status: runtime states such as missing_key, plan_required, rate_limited, and degraded
  • get_api_endpoint_schema: on-demand parameter lookup for any tool

These are the authoritative, runtime-facing views from the live server and should be preferred over any static documentation your agent may have cached.


Example MCP call

bash
curl -X POST https://mcp.hiveintelligence.xyz/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "Authorization: Bearer YOUR_HIVE_API_KEY" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "invoke_api_endpoint",
      "arguments": {
        "endpoint_name": "get_price",
        "args": {
          "ids": "bitcoin",
          "vs_currencies": "usd"
        }
      }
    },
    "id": 1
  }'

Example client configuration

Claude Desktop

Claude Desktop adds remote MCP servers through Settings → Connectors → Add custom connector. Do not paste a url+headers block into claude_desktop_config.json; that file is for stdio servers.

text
Name: Hive
URL: https://mcp.hiveintelligence.xyz/mcp
Authentication: Bearer token
Token: YOUR_HIVE_API_KEY

Cursor

json
{
  "mcpServers": {
    "hive": {
      "url": "https://mcp.hiveintelligence.xyz/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_HIVE_API_KEY"
      }
    }
  }
}

Claude Code

bash
claude mcp add --transport http hive "https://mcp.hiveintelligence.xyz/mcp" \
  --header "Authorization: Bearer YOUR_HIVE_API_KEY"