AI Agents (MCP)

SyncContact ships a remote MCP server — the Model Context Protocol standard AI agents use to work with external tools. Connect an agent once and it can search and update your team's address books — always limited to one workspace and to the scopes you grant.

Server URL:

https://api.synccontact.com/mcp

What agents can do

Area	Tools	Scopes
Contacts	list addressbooks, free-text search, get / create / update contacts	contacts.read, contacts.write

Every tool is workspace-scoped: an agent can never see another workspace's data. Writes land in the contact's history like any other edit, so you can always see what the agent changed — and you can revoke its access at any moment.

Option 1 — connect with a token (Claude Code, Cursor, custom agents)

1. In SyncContact, open your workspace → the AI agents tab.
2. Click New agent token, pick the scopes, and copy the token (sct_…) — it is shown only once.
3. Add the server to your agent:

Claude Code

claude mcp add --transport http synccontact https://api.synccontact.com/mcp \
  --header "Authorization: Bearer sct_…your-token…"

Cursor / VS Code (mcpServers config)

{
  "mcpServers": {
    "synccontact": {
      "type": "http",
      "url": "https://api.synccontact.com/mcp",
      "headers": { "Authorization": "Bearer sct_…your-token…" }
    }
  }
}

Claude API (MCP connector)

{
  "mcp_servers": [
    {
      "type": "url",
      "url": "https://api.synccontact.com/mcp",
      "name": "synccontact",
      "authorization_token": "sct_…your-token…"
    }
  ]
}

Option 2 — connect with OAuth (claude.ai, ChatGPT)

Connectors that manage sign-in themselves only need the server URL:

1. Add a custom connector with https://api.synccontact.com/mcp.
2. The connector opens a SyncContact consent page — sign in, pick the workspace, and choose which scopes to grant.
3. Done. The connection appears in the workspace's AI agents tab as its own token, revocable like any other.

Only workspace owners and admins can authorize agents.

Good to know

• Start read-only. Grant contacts.read first; add contacts.write when you trust the workflow.
• Writes are real edits. An agent with contacts.write changes the same shared address books your team uses — every change is attributed in contact history.
• Revoking is instant. Revoke a token in the AI agents tab and the agent's access dies with it.
• Contact data is untrusted input for your agent. A note or a name field could contain text that tries to steer an AI agent; prefer agents that confirm actions with you before writing.