What is the Convic MCP server?

It's an implementation of the Model Context Protocol that exposes Convic's WhatsApp messaging, contact search, template, and chatbot-building tools to AI clients like Claude, so an AI agent can take real actions in your WhatsApp workspace.

Which AI clients can connect to Convic over MCP?

Any MCP-capable client can connect, including claude.ai, Claude Desktop, Claude Code, and custom applications built on the Anthropic API using the mcp_servers parameter.

Is the Convic MCP connection secure?

The MCP endpoint is scoped to a single workspace, respects the same role-based permissions as the Convic dashboard, logs every action to the audit trail, and can be revoked or rotated at any time from Settings.

Convic

Chatbot Builder ← Back to home Connect your workspace

Model Context ProtocolWorks with Claude & any MCP clientOfficial Convic MCP Server

Give your AI agent a WhatsApp number.

Connect Convic's MCP server to Claude — in claude.ai, Claude Desktop, Claude Code, or your own app on the Anthropic API — and let your AI agent send messages, triage conversations, search contacts, launch campaigns, and build entire chatbot flows. No dashboard required. Just plain language, in, and WhatsApp actions, out.

Get your MCP connection URL See the tool catalog

MCP tools exposed today

1 URL

to connect any MCP client

Scoped

to your workspace only

Live

reads and writes — nothing cached

In plain terms

MCP is how an AI agent gets hands, not just an opinion.

The Model Context Protocol is an open standard that lets an AI assistant call real tools — not guess, not hallucinate a response, but actually reach into a live system and do something. Convic exposes its WhatsApp platform as an MCP server, so any MCP-capable client can read your conversations, send real messages, and manage real chatbot flows on your behalf.

You stay in control: every tool acts only inside your connected workspace, and every action is the same one your team could take from the Convic dashboard — just triggered by a sentence instead of a click.

Claude.ai

chat

Claude Desktop

app

Claude Code

cli

Your app

API

Model Context Protocol

Convic MCP Server

your WhatsApp workspace

Ten tools, one connection

Everything your AI agent can do.

Each tool is scoped to a specific job — no single call can do more than a human on your team could do from the dashboard.

send_whatsapp

Messaging

Send a text or media message to any contact — finds or creates the conversation automatically.

send_template

Messaging

Send an approved template outside the 24-hour window, with variable substitution.

list_conversations

Inbox

Pull open conversations, filter by the 24-hour reply window, or search by contact.

search_contacts

CRM

Find contacts by name, phone, email, company, tag, or status.

list_templates

Campaigns

Browse approved, pending, or rejected templates by category.

upload_media

Media

create_chatbot

Automation

Build a new legacy or branching chain — stages, options, validation, and API calls included.

update_chatbot

Automation

Edit an existing chain's stages, triggers, or status without starting from scratch.

get_chatbots

Automation

List every chain in the workspace, or fetch one chain's full structure by ID.

analytics

Insights

Pull completion stats and per-stage drop-off for any chatbot flow.

What it feels like

You ask. It acts. You stay in the loop.

The agent narrates which tool it's calling and what it found, then asks before anything irreversible goes out — the same judgment you'd expect from a careful teammate.

Claude + Convic MCP

Connected to your workspace

Any conversations still open from the last 24 hours? Give me the top 3 and draft replies.

Called list_conversations(onlyOpenWindow: true, limit: 3)

3 open chats — Ramesh (delivery delay), Priya (refund status), Aakash (won't process payment). Want me to send drafted replies to all three, or review them first?

Architecture

One connection, a clean chain of custody.

Step 1

Your AI client

claude.ai, Claude Desktop, Claude Code, or a custom app on the Anthropic API

Natural language in

Tool calls out

Step 2

MCP transport

The open protocol that carries tool calls and results between client and server

Standardized tool schema

Auth handled per connection

Step 3

Convic MCP server

Exposes messaging, CRM, and chatbot-building tools scoped to your workspace

10 tools available

Same permissions as your dashboard login

Step 4

WhatsApp Cloud API

Where every action actually lands — real messages, real contacts, real chains

Meta Cloud API

Your business number

Setup

Connected in under five minutes.

No new account, no separate login — it uses the workspace you already have.

Get your MCP URL

Get the URL

Step 1

Copy your MCP server URL

From Settings → Integrations, generate a workspace-scoped MCP endpoint. It's unique to your account and can be revoked at any time.

Add the connector

Step 2

Paste it into your MCP client

In claude.ai or Claude Desktop, add it under Connectors. In Claude Code, register it as an MCP server. In your own app, pass it via the mcp_servers parameter on the Anthropic API.

Authorize

Step 3

Confirm the connection

Approve the connection once — your AI client now sees Convic's tools alongside anything else it's connected to.

Prompt

Step 4

Just ask

"Show me today's unread chats" or "build a 4-step feedback flow" — the agent picks the right tool and asks before sending anything irreversible.

For developers

Wire it into your own product.

Pass Convic's MCP endpoint straight into an Anthropic API call and Claude gets access to your WhatsApp tools inside your own application — a support console, an internal ops bot, anything.

request.js

await fetch("https://api.anthropic.com/v1/messages", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    model: "claude-sonnet-4-6",
    max_tokens: 1000,
    messages: [
      { role: "user", content: "Reply to Ramesh about his delayed order." }
    ],
    mcp_servers: [
      {
        type: "url",
        url: "https://convic.in/api/mcp",
        name: "convic-mcp"
      }
    ]
  })
});

Security & control

Agentic. Not unsupervised.

Workspace-scoped

The MCP endpoint only sees the workspace it was generated for — never another account, never another number.

Revoke anytime

Rotate or disable your MCP URL from Settings instantly. Every connected client loses access immediately.

Same permission model

Tool calls respect the exact same role-based permissions your dashboard already enforces.

Full audit trail

Every message sent, contact searched, or chain edited through MCP is logged like any other action.

Confirm before it's irreversible

Well-behaved MCP clients like Claude ask before broadcasting or deleting — you're always the last check.

Meta Terms compliant

Every tool call still runs through Convic's own Meta-compliant messaging layer — nothing bypasses policy.

Get started

Let your AI agent take the first shift.

Connect Convic to Claude and hand off the repetitive parts of WhatsApp — while you keep every decision that matters.

Connect via MCP Explore the chatbot builder

No new account needed·Revoke anytime·Works with any MCP client