Best fit
Founders and teams that want Codex, Claude Code or another MCP client to operate a real omnichannel customer workspace without exposing platform credentials.
Remote MCP · 20 workspace tools
One MCP server for every customer conversation
Give Codex, Claude Code or another AI client tenant-scoped tools to inspect your live workspace, configure assistants while paused, run private tests and control activation across every channel DMtoLead connects.
DMtoLead remote operator20 scoped tools
Instruction
Set up an appointment assistant. Keep every customer reply paused until I approve the private test.
- 01Read only
workspace.statusInspect the live workspace and its next safe step.
- 02Paused
setup.choose_outcomeConfigure the job without starting customer replies.
- 03Private
assistants.test_privateGenerate one customer-invisible answer for review.
- 04Locked
assistants.activateWait for separately authorized, exact approval.
Setup can finish without sending a single customer reply.
20tenant-scoped MCP tools
3separate OAuth permissions
First-party product signal
Meta Tech ProviderShopify App StoreWhat it handles
Your AI gets useful tools. You keep the dangerous switches.
MCP turns a general AI client into a workspace operator without collapsing inspection, configuration and live customer replies into one permission.- 01
One live workspace, not five brittle integrations
The MCP client works through DMtoLead's shared operating layer for Instagram, WhatsApp, Messenger, Telegram and website conversations.
- 02
Configure in private before anything goes live
An AI client can inspect readiness, create and teach an assistant, then run a private test while customer replies remain paused.
- 03
Tools reach the AI. Provider credentials do not
Meta, Telegram, Shopify and other provider tokens remain inside DMtoLead behind tenant scope, OAuth permissions and audited tool calls.
Product facts
The exact permission surface.
- 20 MCP tools
- Inspect status, channels, activity, customers, usage and knowledge; configure assistants; test privately; explicitly activate or pause.
- 3 permission scopes
- dmtolead.read, dmtolead.setup and dmtolead.activate keep inspection, paused configuration and live control separately authorized.
- OAuth 2.1 boundary
- Authorization Code, S256 PKCE, dynamic client registration and rotating refresh tokens protect each remote client connection.
01
02
03
Honest fit
Know where it belongs.
Not the best fit
A raw proxy that hands provider tokens to an agent or lets a prompt silently start customer-facing automation without a separate activation permission.
Why this matters
Your AI already reasons. MCP gives it a safe operating surface
Instead of teaching an AI client five provider APIs, give it one tenant-scoped DMtoLead connection. It can inspect the business, prepare an assistant, prove the behavior privately and stop at the boundary where your approval matters.
The safe operating loop
The AI connects. Your approval boundary stays intact.
- 01
Connect
Authorize the AI client with only the workspace scopes it needs.
- 02
Prepare privately
Inspect, configure and test while automatic replies stay paused.
- 03
Approve deliberately
Activation requires its own scope and explicit confirmation; pause remains immediate.
A closer look
A remote MCP server built for real customer operations
A remote MCP server for Codex and Claude Code
Connect a compatible remote MCP client once, complete OAuth consent and let it discover DMtoLead's twenty structured tools. The client can inspect live workspace state and choose the next safe setup step without scraping dashboards or inventing API calls.
A WhatsApp MCP server without exposing WhatsApp credentials
The AI client can inspect the DMtoLead workspace and configure the assistant serving WhatsApp, while WhatsApp Business credentials remain inside DMtoLead. The same permission boundary applies when the workspace uses QR coexistence or the Cloud API.
Instagram MCP access inside the same omnichannel workspace
Instagram messages and connected channel state live beside the rest of the workspace. Your MCP client operates the assistant and its knowledge layer rather than receiving the Meta token or becoming another channel-specific bot project.
Shopify context through the customer-support MCP layer
Connected Shopify product information can ground customer answers through DMtoLead. The MCP client can inspect knowledge readiness and privately test the assistant without being handed the store's provider credential.
Why read, setup and activate are separate permissions
Reading workspace state is lower risk than changing setup, and changing setup is lower risk than starting live replies. DMtoLead reflects that difference with three OAuth scopes instead of treating every tool call as equally trusted.
What the twenty MCP tools actually cover
The tool surface covers workspace status, channels, recent activity, customers, usage, knowledge search, setup planning, assistant configuration, private tests, acceptance checks, explicit activation and confirmed pause. Every result stays workspace-scoped and carries an audit reference.
Clear answers
Before you let an AI client operate.
01What is a customer support MCP server?+
It is a Model Context Protocol endpoint that gives a compatible AI client structured tools for customer-support work. DMtoLead exposes workspace inspection, paused setup, private testing and explicit reply controls rather than raw provider APIs.
02Can I use the DMtoLead MCP server with Codex?+
Yes. Codex can connect to the remote DMtoLead MCP endpoint through OAuth, discover the available tools and operate within the permissions you approve.
03Does it work with WhatsApp, Instagram and Shopify?+
The MCP connection operates the DMtoLead workspace where those channels and data sources are connected. It does not expose their provider credentials to the AI client.
04Can an MCP client start replying to customers by itself?+
Not through setup permission alone. Live replies require the separate dmtolead.activate scope and an explicit activation confirmation. A confirmed pause is available immediately.
05How many tools does the DMtoLead MCP server expose?+
It currently exposes twenty tools covering workspace reads, setup, private testing, acceptance, activation and pause.
06Are my Meta or Shopify tokens sent to the MCP client?+
No. Provider credentials stay inside DMtoLead. The client receives only tenant-scoped MCP capabilities authorized through OAuth.
07Can a person take over the conversation?+
Yes. Your team can read the conversation, reply manually and take control with the existing context preserved. Manual team replies do not use credits.
08What does the free plan include?+
The Free plan includes 100 monthly credits and does not require a credit card. One automated reply uses one credit. Incoming messages and manual team replies are free.
Your best reply, while they still care