Integrations

Use this page to choose how Hive connects to an agent or application. Start with the managed remote MCP endpoint for supported AI clients, use REST or the TypeScript adapter when your backend owns execution, and use the CLI for local setup, catalog inspection, and scripted calls.

AI clients with native MCP support

Hive ships dedicated install guides for the AI clients and API paths it actively supports. Every guide starts from the same managed endpoint and then shows the field names that client expects:

Cursor, Windsurf, VS Code, Codex CLI, Gemini CLI, Claude Desktop, Claude Code, and OpenAI Responses API each express that contract differently. Use the dedicated guide for the exact config, then verify the connection with a live price or security query. The root MCP endpoint uses tools/list for compact discovery and routing; the full live catalog is available through hive://tools, category endpoints, and GET /api/v1/tools.

• Claude Desktop: Anthropic's macOS/Windows desktop app
• Claude Code: Anthropic's terminal-native coding agent
• Cursor: AI code editor with native MCP support
• Windsurf: Codeium's AI-native IDE with the Cascade agent
• VS Code: Visual Studio Code with GitHub Copilot Chat
• OpenAI Responses API: server-side remote MCP integration, plus ChatGPT app beta auth notes
• Codex CLI: OpenAI's terminal-native coding agent
• Gemini CLI: Google's open-source AI coding agent
• All clients index: pick your client for the full setup walk-through

For local CLI credentials and supported local agent resources, run npx -y -p hive-intelligence@latest hive init --all --browser. For MCP in an AI client, use the dedicated install guide for that client so the transport and auth shape match what the client supports.

Agent frameworks

For multi-step agent workflows such as research, monitoring, alerting, and security review, choose the surface that matches where execution lives. Use the managed MCP endpoint when the framework can connect to remote MCP, use the REST API when your backend controls tool execution, and use the published TypeScript adapter (npm install hive-mcp-client) when you want typed MCP discovery and invocation helpers in Node.

• LangChain & CrewAI: wire Hive tools as LangChain Tool objects or CrewAI agent tools
• REST API tutorial: POST /api/v1/execute with { "tool": "...", "args": { ... } }
• MCP integration tutorial: connect supported and header-capable MCP agents
• CLI tutorial: terminal workflows for tool discovery and execution

Eliza (ai16z) and other agent-character frameworks consume Hive through the same remote MCP or REST contracts. Point the agent's MCP config at https://mcp.hiveintelligence.xyz/mcp, set the auth header, then use Hive discovery resources or REST tool discovery before hard-coding tool names.

Language integrations

For applications embedding Hive into broader products (dashboards, mobile apps, server-side pipelines):

• JavaScript / TypeScript: REST guide plus official hive-mcp-client adapter published on npm at v0.1.5 (npm install hive-mcp-client); source is public at github.com/hive-intel/hive-sdk under client/.
• Python integration: REST examples for scripts and services. A community hive-intelligence package exists on PyPI, but production workloads should call the REST API directly until an official Python SDK ships.
• Go integration: REST quickstart with net/http
• Java integration: REST quickstart with java.net.http.HttpClient
• Rust integration: REST quickstart with reqwest

Other languages can call POST /api/v1/execute with the same auth header and the same REST execution contract.

Integration contract

Preserve these properties in any Hive integration:

1. One auth model. Calls use Authorization: Bearer YOUR_HIVE_API_KEY across remote MCP, REST, CLI, and the TypeScript adapter.
2. Runtime discovery. Agents should discover tools with MCP tools/list, hive://tools, hive://toolsets, category endpoints, or authenticated REST GET /api/v1/tools before hard-coding tool names.
3. Schema-first execution. Inspect schemas before execution, then call exact tools with bounded, schema-valid arguments.
4. Provider-aware answers. User-visible answers should include source or provider context, freshness, runtime status, and any degraded state.
5. Safe failure handling. Treat 401, 403, 429, upstream provider failures, and degraded runtime status as explicit application states.

See the comparisons hub for side-by-side breakdowns of Hive vs CoinGecko MCP, Moralis Cortex, Alchemy, Codex, Birdeye, Helius, QuickNode, DexScreener, and CoinMarketCap.

Production integration checklist

Before exposing Hive results to users:

• Store Hive API keys in server-side environment variables or a secret manager.
• Verify one live tool call from the same client, framework, or backend that will run in production.
• Use discovery and schema lookup before execution instead of relying on stale local tool lists.
• Bound tool calls with limit, cursor, page, per_page, offset, date ranges, or other schema-supported controls.
• Surface provider/source, fetched_at, runtime status, and request errors in user-visible answers or logs.
• Handle 401, 403, 429, provider errors, unavailable providers, and degraded runtime status without hiding the failure.
• Run token security and risk checks before workflows that could lead a user to sign, swap, or size a position.

Related

• Quick Start: run your first Hive request via REST, MCP, or CLI
• API Integration Guide: auth, discovery, transport patterns
• Tools Reference: full tool catalog across 10 categories
• Glossary: definitions for MCP, tool calls, agent skills, and crypto terminology
• Comparisons: Hive vs CoinGecko, Moralis, Alchemy, Codex, Birdeye, Helius, and more