Coevera AI Connectors v2.0.9

Service enablement guide for the Coevera MCP Server and the Automation Hub connectors that link a customer's CRM to AI assistants such as Claude and ChatGPT.

Who this guide is for. This page is written for our Global Service team as source material for the final customer-facing documentation. It favours plain language and business context over engineering detail. The full technical reference for developers lives at /doc.

What it is

A secure bridge that lets an AI assistant read and update a customer's Coevera CRM in plain conversation.

How customers get it

Three connectors in the Automation Hub: verified Claude and ChatGPT connectors, plus a configurable Custom connector.

Why it matters

No exports, no copy-paste. The customer asks; the AI answers and acts on live CRM data, with their own permissions.

Overview

The Coevera MCP Server connects a customer's CRM to AI assistants. Instead of exporting spreadsheets and pasting them into a chat window, the customer's chosen assistant connects directly to their Coevera space and can search, summarise, create, and update CRM records on request.

It is built on the Model Context Protocol (MCP) — an open industry standard for connecting AI assistants to external tools and data. Because it is a standard, the same server works with Claude, ChatGPT, and other MCP-compatible assistants. The customer never sees the protocol; they just enable a connector and start asking questions.

Customers reach these connectors from CRM Space Administration, under Automation Hub. There are three of them, described next.

The Three Automation Hub Connectors

ConnectorStatusConfigurationIntended audience
Claude ConnectorVerifiedNo — enable & connectEnd customers using Claude (claude.ai, Claude Desktop, or the Claude apps).
ChatGPT ConnectorVerifiedNo — enable & connectEnd customers using ChatGPT.
Custom MCP ConnectorCustomYes — connection settings requiredInternal development and testing, custom in-house assistants, and approved third-party AI agents that speak the MCP standard.

The two verified connectors (Claude and ChatGPT) are officially approved and listed in the Coevera app catalogue. They have no configuration options — the customer simply enables the connector and connects their AI account. The Custom connector is configurable and is meant for development and third-party agents.

Claude Connector Official, verified — for Anthropic Claude
Verified connector BETA

A pre-approved, verified connector published in the Coevera app catalogue. It is dedicated to Anthropic Claude and ships with everything pre-configured. There are no settings to fill in — the customer simply enables it and connects their Claude account.

Who it's for: End customers using Claude (claude.ai, Claude Desktop, or the Claude apps).

AI
Claude Connector
Productivity BETA
As it appears in Automation Hub › Addons
ChatGPT Connector Official, verified — for OpenAI ChatGPT
Verified connector BETA

A pre-approved, verified connector published in the Coevera app catalogue. It is dedicated to OpenAI ChatGPT and, like the Claude connector, has no configuration options. The customer enables it and connects their ChatGPT account.

Who it's for: End customers using ChatGPT.

AI
ChatGPT Connector
Productivity BETA
As it appears in Automation Hub › Addons
Custom MCP Connector Configurable — for development & third-party AI agents
Configurable BETA

A flexible, configurable connector for cases the two verified connectors do not cover: development and QA, bespoke integrations, or connecting a third-party AI agent. Because it can point at any compatible MCP endpoint, it exposes connection settings (covered in detail below) and is intended for technical users rather than a one-click customer experience.

Who it's for: Internal development and testing, custom in-house assistants, and approved third-party AI agents that speak the MCP standard.

AI
Custom MCP Connector
MCP Connector BETA
Running
As it appears in Automation Hub › Addons

Custom MCP Connector — Configuration

Unlike the verified connectors, the Custom MCP Connector opens a configuration dialog. It is used when a connection must point at a specific MCP server — for example an internal development or staging environment, or an approved third-party AI agent. The dialog collects the following:

FieldWhat it means
Server URLThe address of the MCP server the connector talks to, ending in /mcp (for example a staging or production Coevera MCP endpoint). This is what makes the connector "custom" — the verified connectors have this baked in.
Client IDThe public identifier of the OAuth application the connector authenticates as. Issued when the connector is registered with the CRM.
Client Secret secretThe confidential password paired with the Client ID. Treated as a secret — it is masked in the dialog and must never be shared in customer-facing documents.
Redirect URLsThe allow-list of addresses the OAuth server may send the user back to after they sign in and approve access. Each entry is added explicitly; requests to any other address are rejected. This is a security control that prevents tokens from being delivered to an untrusted destination.
Coevera MCP Server — Configure ×
Credentials
Close Save

Illustration of the configuration dialog — values shown are placeholders.

Handle with care. The Custom connector is a technical, developer-facing tool. The Client Secret is a credential and must never appear in customer-facing documentation or support tickets. For normal customers, always point to the verified Claude or ChatGPT connector instead.

How a Customer Enables & Connects

Where to find the connectors

CRM › Space Administration › Automation Hub › Addons › locate the connector tile

☍ Coevera
  • General
  • Units, Users & Roles
  • Processes
  • Currencies
  • Products & Prices
  • Common Lists
  • Entities
  • Settings
  • Automation Hub
  • Security
Addons Integrations via Zapier Analytics Logs
AI
Claude Connector
Productivity BETA
AI
ChatGPT Connector
Productivity BETA
AI
Custom MCP Connector
MCP Connector BETA
Running

Verified connector (Claude or ChatGPT)

  1. Open Space Administration → Automation Hub and select the Addons tab.
  2. Find the Claude Connector or ChatGPT Connector tile and open it.
  3. Click Enable — there is nothing to configure.
  4. In the AI assistant (Claude or ChatGPT), add the Coevera connector and start the connection.
  5. The customer is taken to a Coevera sign-in & consent screen, where they log in and approve the access the connector is requesting. After approval they are returned to the assistant — connected and ready.
The connection uses a standard, secure OAuth sign-in. The customer authenticates with their own Coevera credentials, so the assistant only ever sees what that user is allowed to see.

Custom connector

  1. Open the Custom MCP Connector tile in the Automation Hub.
  2. Enter the Server URL, Client ID, Client Secret, and the allowed Redirect URLs (see the configuration section above).
  3. Save the configuration and connect the AI agent, completing the same OAuth sign-in & consent step.

Because it requires connection details and credentials, the Custom connector is for technical users — not part of the standard customer self-service flow.

What the AI Can Do Once Connected

Everything below respects the signed-in user's permissions. Rows marked reads data only look things up; rows marked changes data create or modify CRM records.

CapabilityEffectExample
Ask questions about CRM datareads data“Which accounts in Germany have open opportunities over €50k?” The assistant searches and filters live CRM records and answers in natural language — no manual exports or report building.
Summarise and segmentreads dataSummarise this quarter's open pipeline, group leads by source, or list every contact at an account. The assistant reads records and synthesises them on demand.
Discover the data modelreads dataThe assistant can list the entity types a user can see (Accounts, Contacts, Leads, Opportunities, Tasks, custom entities, …) and inspect their fields, so it always works with the customer's real schema — including custom fields.
Create recordschanges data“Create a follow-up task for this contact next Tuesday.” The assistant creates Contacts, Tasks, Appointments, Opportunities, and more, with the user's permissions.
Update recordschanges dataChange a deal's close date, update a contact's phone number, or correct an account field — the assistant updates only the fields it is told to.
Attach documentschanges data“Upload this signed contract to the account's Documents folder.” The assistant can file documents the user shares with it (PDFs, images, spreadsheets, …) into a CRM document folder, where they appear like any other uploaded file. Uploads happen through a short-lived, single-use link, and the file is stored under the user's own identity and permissions. Note: depending on the client application (Claude or ChatGPT), the domain *.coevera.com may need to be allowed in the app's security settings before file uploads work — see the app preferences.
Run business operationschanges dataBeyond simple edits, the assistant can perform real sales operations: move or archive batches of opportunities, qualify leads into opportunities, reassign ownership, and complete pipeline-step checklist requirements — each triggering the CRM's own business logic.
The assistant works across the customer's CRM objects — Accounts, Contacts, Leads, Opportunities, Tasks, Appointments, Products, and any custom entities — using the customer's own field names and configuration.

Security & Data Handling (in plain terms)

  • The customer signs in themselves. Connecting uses OAuth — the customer logs in with their Coevera account and explicitly approves access. We never ask for or store their password.
  • Permissions are respected. The assistant can only see and change what the signed-in user is allowed to. It inherits that user's CRM permissions exactly.
  • Each customer is isolated. Every request is tied to the customer's own CRM space (tenant), so one customer's assistant can never reach another customer's data.
  • Access can be revoked. Because access is a granted authorisation, the customer or an administrator can disable the connector or revoke access at any time.
  • Data stays on the platform. The connector reads and writes the customer's CRM through official Coevera APIs. It does not copy CRM data to third-party storage.
  • Fair-use protection. Built-in rate limiting protects the CRM from overload, keeping the service responsive for everyone in the space.
For the customer documentation: emphasise that the customer is always in control — they choose which assistant to connect, they sign in with their own account, they approve exactly what is shared, and they can disconnect whenever they wish.

FAQ

Is this only for Claude and ChatGPT?
Those are the two officially verified connectors today. Any other MCP-compatible assistant can be connected through the Custom connector, which is aimed at developers and approved third-party agents.
Do the verified connectors need any setup?
No. They are pre-approved and pre-configured. The customer enables the connector and signs in — there are no fields to fill in.
Can the AI change CRM data, or only read it?
Both, within the user's permissions. It can answer questions (read) and also create or update records and run business operations (write) when asked.
Can the AI upload files into the CRM?
Yes. When a user shares a file with the assistant and asks for it to be filed, the assistant stores it in a CRM document folder as a regular document. The file is transferred over a one-time link that expires after a few minutes and cannot be reused, and the upload is done as the signed-in user — so it only works for folders that user is allowed to write to. This requires the documents.full permission, which is part of the standard set the customer approves at sign-in.

One extra step may be needed in the AI app. The file itself is sent by the client application (Claude or ChatGPT) directly to our upload address, and both apps restrict where they are allowed to send data. Depending on the app and its administrator settings, the domain *.coevera.com has to be added to the allowed domains in the app's security / network preferences — otherwise the assistant can prepare the upload but the file never arrives. If a customer reports that “the upload fails” while everything else works, check this setting first.
What does the customer agree to when connecting?
During sign-in they approve a set of access permissions (scopes) covering the CRM areas the connector works with — contacts, accounts, leads, opportunities, tasks, and so on. The full list is in the Glossary below.
Are these connectors generally available?
The connectors are currently marked BETA in the Automation Hub. Reflect that status in customer-facing materials until it changes.
Where can engineering find the technical details?
The developer reference — tools, parameters, filter syntax, endpoints, and troubleshooting — is served at /doc.

Glossary & Reference

MCP
Model Context Protocol — the open standard for connecting AI assistants to tools and data.
Connector
An Automation Hub item that links an AI assistant to the customer's CRM.
Automation Hub
The area of Space Administration where connectors and integrations are managed.
Space
A customer's isolated CRM tenant. All access is scoped to one space.
OAuth
The secure sign-in standard used to grant access without sharing a password.
Scope
A specific permission the customer approves at sign-in (e.g. access to contacts).
CRM API host
crm.coevera.com

Access permissions (scopes) a customer approves

When connecting, the customer grants the connector these CRM permissions:

contacts.fullaccounts.fullleads.fulltasks.fullemails.fullappointments.fullmessages.fullopportunities.fullprojects.readcurrencies.readproducts_and_prices.fullpipelines.readfields_and_forms.readdocuments.fullunits_and_roles.read