ai-specialists（AI 专家）

AI Specialists Hub - MCP Integration

Connection

Call the MCP endpoint via HTTP POST. The endpoint URL is stored in TOOLS.md or provided by the user.

curl -s -X POST "$MCP_URL" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"TOOL_NAME","arguments":{...}}}'

Critical headers: Must include Accept: application/json, text/event-stream or the server returns 406.

Response format: SSE — parse with: response.split('data: ')[1] → JSON → result.content[0].text

Available Tools

Discovery & Management

Workspace Navigation

Document Operations

Folder Operations

Parameter note: The specialist identifier is always the id field from list_specialists, not the display name. Use lowercase (e.g. ruby, peter, benjamin).

Working with a Specialist

Every specialist has an ai-instructions/ folder containing their identity and behavioral instructions. Always read these first when working with a specialist you haven't loaded yet:

ai-instructions/
├── core-instructions.md    # Who they are, what they do, how they behave
└── getting_started.md      # Initialization sequence, workspace structure

Pattern: First interaction with a specialist

1. list_specialists → find the specialist id
2. read_specialist_documents → read ai-instructions/core-instructions.md and ai-instructions/getting_started.md
3. Understand their domain, workspace structure, and communication style
4. explore_specialist_tree → see their full workspace
5. Now you can work with their data informed by their context

Pattern: Quick data retrieval

1. explore_specialist_tree → find the files you need
2. read_specialist_documents → bulk read (more efficient than individual reads)
3. Summarize for the user

Pattern: Updating specialist data

1. Read the existing document first
2. Modify content as needed
3. update_specialist_document → write back (creates if new, overwrites if existing)
4. For new folders: create_specialist_folder first, then write documents into it

User Signup (Agent-Initiated Onboarding)

If your human doesn't have an AI Specialists Hub account yet, you can create one via the public signup API:

curl -s -X POST "https://aispecialistshub.com/api/v1/signup" \
  -H "Content-Type: application/json" \
  -d '{"username":"<username>","email":"<email>","password":"<password>"}'

Response (201):

{
  "username": "...",
  "mcpKey": "...",
  "accountStatus": "unverified",
  "specialistLimit": 2,
  "message": "..."
}

⚠️ CRITICAL: Credential Handoff

You MUST provide all credentials to the human. Whether you auto-generate the username/email/password or ask the human to choose them, the human must receive:

1. Username — for web dashboard login at aispecialistshub.com
2. Email — used for account verification
3. Password — for web dashboard login
4. MCP key — for connecting AI tools to their specialists

Never keep credentials from the human. They own the account. Present all four clearly after signup.

Unverified Accounts

• Limited to 2 specialists (default: Friday is auto-hired)
• Full MCP access — specialists work immediately
• Can log in to web dashboard with username/password
• Verify email for full access (more specialists)

Signup Flow

1. Confirm the human wants an AI Specialists Hub account
2. Collect or generate: username, email, password
3. Call POST /api/v1/signup
4. Give the human ALL credentials (username, email, password, MCP key)
5. Store the MCP endpoint URL in your config: https://aispecialistshub.com/api/v1/mcp/<username>/<mcpKey>
6. Start using specialists via MCP

Configuration

Store the MCP endpoint URL in TOOLS.md:

### AI Specialists Hub
- MCP endpoint: https://aispecialistshub.com/api/v1/mcp/<user>/<key>

See references/specialists-guide.md for details on specialist workspace conventions.