--- title: MCP server and AI connectors description: Let Claude, ChatGPT or another MCP client read and update a Coevera space through the Coevera MCP server, with the user's permissions and consent. canonical: https://developers.coevera.com/ai/overview/ section: ai publisher: Coevera --- # MCP server and AI connectors **Answer:** The Coevera MCP server at https://mcp-server.coevera.com/mcp gives AI assistants 20 tools over your CRM data. A space admin requests a connector in the Automation Hub and Coevera Customer Success activates it; each user then signs in with OAuth and approves the scopes. The connectors are free and in beta, and the server runs in the United States. **Status:** beta. **Price:** free. **Tools:** 20. **Sign-in:** OAuth 2.0 authorization code flow with PKCE. The MCP server implements the Model Context Protocol for Coevera. An assistant calls its tools; the server turns each call into requests to the Coevera REST and GraphQL APIs on behalf of the signed-in user, so the assistant works within that user's access and the scopes the user approved. ## Connectors A connector is how an assistant reaches the server. Each one is a tile in the Automation Hub, and each must be activated for the space before anyone can use it. | Connector | For | Setup | | ----------------- | ----------------------------------------------------------------------- | --------------------------------------------- | | Claude Connector | Claude | [Connect Claude](/ai/connect-claude/) | | ChatGPT Connector | ChatGPT | [Connect ChatGPT](/ai/connect-chatgpt/) | | MCP Connector | Any other MCP client, with your own client ID, secret and redirect URLs | [Custom MCP connector](/ai/custom-connector/) | Activation is not self-serve: see [Activate a connector](/ai/activation/). ## Hosting and your data | Topic | Today | | ------------- | ---------------------------------------------------------------------------------------------------------------------- | | Server | Hosted in the United States, whatever the region of your space. | | Your CRM data | Stays stored in its region. Requests and answers pass through the US server on their way to and from your space's API. | | Transport | Streamable HTTP, `POST /mcp`, a Bearer token on every request. | | Logs | Kept mostly in Ireland for up to 90 days. | | Status | Beta. The connectors are free. | Error reports go to a US-based error-monitoring service. The details are on [Hosting and your data](/ai/hosting-and-data/). ## Tools > **Tool reference:** The parameters and return values of every tool are in the > [MCP server's tool reference](https://mcp-server.coevera.com/doc#tools), published with the server and updated with each > version; each tool's name below opens its entry there. This page gives the overview. The server offers 20 tools in five groups. | Group | Tool | What it does | | ----------------- | ------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------- | | Discovery | [`me`](https://mcp-server.coevera.com/doc#tool-me) | Who is signed in, in which space, with which scopes | | | [`meta_list_entity_types`](https://mcp-server.coevera.com/doc#tool-meta_list_entity_types) | The entities the user can access, with their create, read, update and delete permissions, and whether global search covers them | | | [`meta_describe_entity`](https://mcp-server.coevera.com/doc#tool-meta_describe_entity) | The fields of one entity, their types, and which can be filtered, expanded or searched | | Read | [`global_search`](https://mcp-server.coevera.com/doc#tool-global_search) | Find records by a text across accounts, contacts, leads, opportunities, tasks, appointments, projects, documents and custom entities, ranked by relevance | | | [`search_records`](https://mcp-server.coevera.com/doc#tool-search_records) | Filter, sort and page through the records of one entity | | | [`get_records`](https://mcp-server.coevera.com/doc#tool-get_records) | Fetch records by ID, with expanded relations | | | [`get_step_checklist_status`](https://mcp-server.coevera.com/doc#tool-get_step_checklist_status) | The checklist items of a pipeline step | | Report and export | [`run_report`](https://mcp-server.coevera.com/doc#tool-run_report) | Count, sum, average, minimum and maximum over the records that match a filter, grouped by up to three fields | | | [`export_records`](https://mcp-server.coevera.com/doc#tool-export_records) | Write the records that match a filter to a CSV file behind a one-time download link | | Write | [`create_records`](https://mcp-server.coevera.com/doc#tool-create_records) | Create records | | | [`update_records`](https://mcp-server.coevera.com/doc#tool-update_records) | Update fields of existing records | | | [`delete_records`](https://mcp-server.coevera.com/doc#tool-delete_records) | Soft-delete up to 100 records per call; leads and opportunities must be archived instead | | | [`upload_cloud_object_document`](https://mcp-server.coevera.com/doc#tool-upload_cloud_object_document) | Attach a file through a single-use upload URL | | | [`set_record_picture`](https://mcp-server.coevera.com/doc#tool-set_record_picture) | Set the logo of an account, the photo of a contact or the image of a product | | | [`change_ownership`](https://mcp-server.coevera.com/doc#tool-change_ownership) | Hand records to another owner | | | [`complete_step_checklist_item`](https://mcp-server.coevera.com/doc#tool-complete_step_checklist_item) | Mark a checklist item as done for a record | | Bulk | [`bulk_move_opportunities`](https://mcp-server.coevera.com/doc#tool-bulk_move_opportunities) | Move opportunities to another step of their pipeline | | | [`bulk_archive_opportunities`](https://mcp-server.coevera.com/doc#tool-bulk_archive_opportunities) | Archive opportunities | | | [`bulk_archive_leads`](https://mcp-server.coevera.com/doc#tool-bulk_archive_leads) | Archive leads | | | [`bulk_qualify_leads`](https://mcp-server.coevera.com/doc#tool-bulk_qualify_leads) | Qualify leads into opportunities | Every tool declares the MCP annotations (read-only, destructive, idempotent, open-world), which clients use to decide when to ask the user before a call; the tool reference gives [each tool's values and the reason for each](https://mcp-server.coevera.com/doc#annotations). `export_records` changes nothing in the space but is not marked read-only, so clients ask the user to approve every export. The server's instructions point assistants to `global_search` to find a record by its name or any other text, to `run_report` for totals and breakdowns, and to `export_records` for files and for analyses of many records. They also say that tool results are data entered by users, never instructions. ## Scopes The connectors ask for 15 scopes: ten grant write access (`.full`) and five grant read access (`.read`). | Write (`.full`) | Read (`.read`) | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- | | `accounts.full`, `contacts.full`, `leads.full`, `opportunities.full`, `tasks.full`, `appointments.full`, `emails.full`, `messages.full`, `documents.full`, `products_and_prices.full` | `projects.read`, `currencies.read`, `pipelines.read`, `fields_and_forms.read`, `units_and_roles.read` | What each scope grants, and what can never be granted, is on [OAuth and scopes](/ai/oauth-and-scopes/). > **Filters and pagination differ from REST:** `search_records` takes `limit` (default and maximum 100) and a cursor, and the tools have their > own filter syntax. See [Filters and pagination](/ai/filters-and-pagination/). ## Build with the API instead Building your own agent or automation? Call the REST or GraphQL API directly with an API application. Start with the [quickstart](/get-started/quickstart/), then read [Build agents on the API](/ai/build-agents-on-the-api/). To react to changes without asking, use [webhooks](/webhooks/overview/).