MCP · Streamable HTTP · 43 Tools

AI agents that actually build your community.

Connect Claude, Cursor, or any MCP-compatible AI to HiveLearn in 60 seconds. Your assistant can draft courses, schedule events, answer member questions, and issue certificates — through the same public API you'd use yourself.

Get Startedmcp.hivelearn.app
https://mcp.hivelearn.app·Streamable HTTP

What is MCP?

Model Context Protocol is an open standard for connecting AI assistants to real systems — databases, APIs, file systems, and SaaS products. Think of it as "USB-C for AI": one cable, many devices.

The HiveLearn MCP server exposes 43 tools across 10 domains. Any MCP-compatible assistant (Claude Desktop, Claude Code, Cursor, IDE integrations) can read your community data and take actions on your behalf — with per-key scopes and rate limits.

Setup (60 seconds)

Two transports, same server. Use HTTP when your client supports it; stdio otherwise.

1

Create an API key

In your community: Admin → API Keys → Create Key. Requires a Pro, Scale, or Enterprise community (or a pioneer community). Grant the scopes your agent needs — read:members, write:courses, etc.

2

Choose your client

Claude Desktop (HTTP — recommended)

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

claude_desktop_config.jsonjson
{
  "mcpServers": {
    "hivelearn": {
      "url": "https://mcp.hivelearn.app",
      "headers": {
        "X-Hivelearn-Api-Key": "hl_live_YOUR_KEY_HERE"
      }
    }
  }
}

Claude Code

Terminalbash
# Add the server
claude mcp add hivelearn npx -- -y hivelearn-mcp

# Set your API key
claude mcp set-env hivelearn HIVELEARN_API_KEY=hl_live_YOUR_KEY_HERE

Cursor / other stdio-only clients

Uses the hivelearn-mcp npm shim:

~/.cursor/mcp.jsonjson
{
  "mcpServers": {
    "hivelearn": {
      "command": "npx",
      "args": ["-y", "hivelearn-mcp"],
      "env": {
        "HIVELEARN_API_KEY": "hl_live_YOUR_KEY_HERE"
      }
    }
  }
}
3

Restart your client, then try a prompt

Your assistant will see the HiveLearn tools the moment it connects. Ask for something natural and watch it pick the right tool.

Tool Catalog

Full OpenAPI spec

🏛️ Community

1 tool
  • hivelearn_get_community_me

👥 Members

2 tools
  • hivelearn_list_members
  • hivelearn_get_member

📝 Posts

6 tools
  • hivelearn_list_posts
  • hivelearn_get_post
  • hivelearn_create_post
  • hivelearn_update_post
  • hivelearn_list_post_comments
  • hivelearn_create_post_comment

📅 Events

6 tools
  • hivelearn_list_events
  • hivelearn_get_event
  • hivelearn_create_event
  • hivelearn_update_event
  • hivelearn_list_event_attendees
  • hivelearn_rsvp_event

🎓 Courses

7 tools
  • hivelearn_list_courses
  • hivelearn_get_course
  • hivelearn_create_course
  • hivelearn_update_course
  • hivelearn_publish_course
  • hivelearn_get_course_structure
  • hivelearn_create_course_outline

📚 Modules

3 tools
  • hivelearn_create_module
  • hivelearn_update_module
  • hivelearn_reorder_modules

📖 Lessons

3 tools
  • hivelearn_create_lesson
  • hivelearn_update_lesson
  • hivelearn_update_lesson_content

✅ Quizzes

5 tools
  • hivelearn_list_quizzes
  • hivelearn_get_quiz
  • hivelearn_create_quiz
  • hivelearn_add_quiz_question
  • hivelearn_list_quiz_attempts

🎟️ Enrollments

5 tools
  • hivelearn_list_enrollments
  • hivelearn_enroll_member
  • hivelearn_update_enrollment
  • hivelearn_mark_lesson_complete
  • hivelearn_get_enrollment_progress

🏅 Certificates

4 tools
  • hivelearn_list_certificates
  • hivelearn_get_certificate
  • hivelearn_issue_certificate
  • hivelearn_verify_certificate

No delete_* tools are exposed. Agents can draft, publish, and unpublish — but they can't destroy content.

What You Can Ask

"Draft a 4-module course called "Launch on Product Hunt" with 3 lessons each — but don't publish yet."

calls → create_course_outline

"Show me all pending quiz attempts from the last 30 days, grouped by course."

calls → list_quiz_attempts

"List the top 5 most-RSVP'd events this quarter."

calls → list_events, list_event_attendees

"Find members who enrolled in "Intro to Capacitor" but haven't completed the first lesson yet."

calls → list_enrollments, get_enrollment_progress

"Publish course abc-123 now that all lessons are ready."

calls → publish_course

Security

  • Keys are hashed server-side (SHA-256). Plaintext shown once at creation — rotate or revoke any time.
  • Per-key scopes. Grant only what the agent needs (read:members, write:courses, etc).
  • Per-community tenancy. The key determines the community — an agent cannot read or write another community's data.
  • Rate-limited and logged. Every tool call is tracked per key in your admin panel.
  • No delete tools. The worst an agent can do is unpublish — never permanently remove content.

FAQ

Which plan do I need?

API keys are available on Pro, Scale, Enterprise, and pioneer communities. Any plan can use MCP once keys are issued.

Does this work with ChatGPT or Gemini?

MCP is an open standard — any compliant client works. Claude Desktop, Claude Code, Cursor, and Zed all support it natively today. ChatGPT and Gemini are adding support progressively.

Can I self-host?

Yes — the MCP server is part of the HiveLearn repo (api/mcp.ts on Vercel) and points at your Supabase API. Set HIVELEARN_MCP_URL to your own endpoint when using the stdio shim.

What happens if the agent hits a rate limit?

The MCP client surfaces the error to the model, which typically backs off and retries. Rate limits are per-key, configurable per tier.

Does the agent see member emails?

Only if the key has the read:members scope. Without it, agents see public profile fields only (display name, avatar, role).

Ready to let your AI do the work?

Create your community, generate a key, and plug in your assistant. No credit card, free on every tier with API access.

Get Early Access — FreeView on GitHub