WhatsApp MCP by TimelinesAI

Full-blown MCP server — every workspace action exposed to your AI assistant.

Drive your TimelinesAI WhatsApp inbox from Claude, Cursor, Claude Code, or any MCP-compatible client. List chats, read history, send messages, react, label, assign teammates, check quotas — all 18 tools, all as you, all on the same quota the TimelinesAI UI uses.

---

What is this?

The TimelinesAI MCP server is a remote control for your real WhatsApp inbox inside TimelinesAI. Connect once via OAuth, then your AI assistant gains 18 tools across chat discovery, messaging, triage (labels + assignments), and workspace introspection.

No sandbox. Writes are real and visible to your contacts. Recipients cannot tell whether a message came from the UI or from an AI assistant — it's all just you.

Requirements

• A working TimelinesAI production account with at least one connected WhatsApp account in your workspace. Without a connected WA, most tools will return empty or error — the MCP is a remote control for your real inbox, not a sandbox.
• Node.js 18+ locally (Cursor / Claude Desktop / Claude Code use npx to launch the mcp-remote proxy).

Don't have a TimelinesAI account yet? Sign up free →

---

Quick setup (~1 min)

Claude Code (CLI)

claude mcp add timelinesai -- npx -y mcp-remote@latest https://mcp.services.timelines.ai/mcp --host 127.0.0.1

Restart Claude Code. A browser tab will open on first use — log in with your TimelinesAI account to authorize. Done.

Cursor

Edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "timelinesai": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@latest",
        "https://mcp.services.timelines.ai/mcp",
        "--host", "127.0.0.1"
      ]
    }
  }
}

Restart Cursor.

Claude Desktop

Open Settings → Developer → Edit Config and add:

{
  "mcpServers": {
    "timelinesai": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@latest",
        "https://mcp.services.timelines.ai/mcp",
        "--host", "127.0.0.1"
      ]
    }
  }
}

Fully quit Claude Desktop (not just close the window) and reopen. A browser tab will open for OAuth on first use.

See examples/ for ready-to-copy config snippets per client.

---

First thing to try

After your client restarts, ask it:

"What tools does the timelinesai MCP server expose? Group them by purpose and give me one example prompt per group."

That gives you a personalized tour of what's available without reading any docs.

---

Tool catalog (18 tools)

Workspace meta — 3 tools

Read-only introspection. Cheap to call.

Tool	Purpose
workspace_quotas	Current plan, seats, messaging + API call quotas, billing period
workspace_whatsapp_accounts	Connected WhatsApp accounts (id, phone, owner, status)
workspace_team	Teammates, roles, invitation status, account bindings

Chat discovery + inspection — 5 tools

Browse and read. No writes.

Tool	Purpose
list_chats	Filter chats by status, labels, assignee, group/direct, phone, name, dates, WA account — paginated 50/page
chat_details	Full metadata for a single chat
chat_history	Conversation context window around a specific message
get_chat_messages	Page through messages in a chat
message_details	Inspect a single message

Messaging writes — 4 tools

Quota-consuming. Same monthly messaging budget as the TimelinesAI UI.

Tool	Purpose
chat_send_message	Send a message (text, attachment, or reply) in an existing chat
whatsapp_account_send_message	Send a message to any phone via a specific WA account (creates chat if none exists — cold send)
message_reply	Threaded reply to a specific message
message_react	Set or clear an emoji reaction on a message

Chat mutations — 6 tools

State + triage operations on chats. Idempotent label and assign ops.

Tool	Purpose
chat_open	Reopen a closed chat
chat_close	Close a chat
chat_assign	Assign chat to a teammate by email (use workspace_team to discover emails)
chat_unassign	Unassign current responsible teammate
chat_set_label	Add a label to a chat
chat_remove_label	Remove a label from a chat

→ Full schemas and example prompts: docs/tools.md

---

How authentication works

The server uses OAuth via the mcp-remote proxy. On first connection, your browser opens to mcp.services.timelines.ai — log in to your TimelinesAI account to authorize. Token is cached locally by mcp-remote; subsequent sessions reconnect without prompting.

On machines that already have an active TimelinesAI session (e.g. you're logged in to TimelinesAI in your default browser, or you've previously authorized the MCP via the Claude.ai integration), authorization may complete silently with no browser prompt.

Every action runs as you — your role's permissions in TimelinesAI apply to MCP calls exactly as they apply to UI actions.

---

Safety + production behavior

⚠️ This MCP server controls a real WhatsApp inbox. Read this section before sending writes.

• Messages sent via MCP are indistinguishable from messages sent in the UI. Recipients cannot tell.
• Writes are real and immediate. No undo. Confirm before bulk actions.
• Quota is shared with UI usage — MCP messages draw from the same monthly messaging budget. Call workspace_quotas to see headroom.
• No sandbox mode. Practice on your own number first if you're unsure.
• Without a connected WhatsApp account, most tools return empty results or errors. Connect a WA account in TimelinesAI first.

---

Get started

1. Sign up for TimelinesAI →
2. Connect a WhatsApp account in the TimelinesAI dashboard.
3. Pick your client above (Claude Code, Cursor, or Claude Desktop) and paste the config.
4. Restart, authorize, ask your assistant what it can do.

---

Links

• TimelinesAI — timelines.ai
• Sign up — app.timelines.ai/register
• MCP server endpoint — https://mcp.services.timelines.ai/mcp
• Model Context Protocol — modelcontextprotocol.io
• mcp-remote proxy — npm
• Issues / feedback — open an issue in this repo

---

License

MIT © 2026 Initech Software / TimelinesAI