---
name: hive-mcp
description: "Use this skill when installing, configuring, or debugging Hive MCP in an AI client such as Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, Gemini CLI, the OpenAI Responses API, or Codex CLI."
license: MIT
metadata:
  package: "@hiveintelligence/agent-skills"
  category: "setup"
  requires_network: "true"
version: 1.0.0
---

# Install Hive MCP

Connect Hive's hosted remote MCP endpoint to Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, OpenAI Responses API, Codex CLI, and Gemini CLI.

## Trigger

Use this skill when installing, configuring, or debugging Hive MCP in an AI client such as Claude Desktop, Claude Code, Cursor, Windsurf, VS Code, Gemini CLI, the OpenAI Responses API, or Codex CLI.

## Use when

- install hive
- connect mcp
- configure client

## Source

Canonical source path: `agent-skills/packs/hive-mcp/SKILL.md`

## Task toolsets

- `search_discovery` — Route ambiguous requests to exact task toolsets, categories, tools, schemas, and entity identifiers. Required identifiers: natural-language query.

## Procedure

1. Read `hive://toolsets` or call `search_tools` to confirm the task.
2. Ask for any missing required identifier before execution.
3. Call `get_api_endpoint_schema` before invoking an exact endpoint.
4. Call `invoke_api_endpoint` with schema-valid arguments.
5. Report freshness, source, warnings, and runtime status.

## Runtime status handling

Hive uses `ok`, `missing_key`, `plan_required`, `rate_limited`, `degraded`, `failing`.
A non-`ok` state means the tool exists but the runtime cannot complete that call
right now. Do not hide or remove the tool.

## Source of truth

The executable MCP skill pack ships with the Hive MCP package. This public file
is a website mirror for agent discovery. Runtime metadata is available at
`https://mcp.hiveintelligence.xyz/mcp` via `hive://skills`. Installable skills are mirrored at
https://github.com/hive-intel/hive-skills; install the full pack with
`npx skills add hive-intel/hive-skills`.