Kommo CRM

What the Kommo CRM MCP server does

The Kommo CRM MCP server gives AI agents access to common Kommo CRM operations through the Model Context Protocol. It works with Kommo, formerly known as amoCRM, and targets workflows involving sales records and related customer data.

The available operations cover lead retrieval and management, including creation, updates, deletion, bulk updates, and stage changes. Agents can also work with contacts, companies, pipelines, stages, tags, tasks, notes, chat templates, and messages. Custom-field tools support field types such as text, select, multiselect, and date. An advanced operation can create a lead, contact, and company together.

Reach for this server when an MCP client needs to inspect CRM data or carry out routine sales-operations changes without building a separate Kommo API client. Example requests include finding leads in a pipeline, creating a lead from contact details, moving a lead to a new stage, or listing overdue tasks.

How it works

The Kommo CRM MCP server is implemented in Python and communicates with Kommo through its HTTP API. It uses the stdio transport, so an MCP client starts the local command and exchanges tool calls with the process. The package provides the kommo-mcp entry point and can also be launched with python -m kommo_mcp.

Authentication uses Kommo OAuth v4. A setup script opens a browser for sign-in and authorization, then saves the resulting tokens. The client refreshes tokens automatically. Requests are limited to five per second, failed requests use exponential backoff with jitter, and selected pipeline, stage, and custom-field responses are cached for 10 minutes.

Setup and configuration

Create a Kommo integration under Settings → Integrations, copy its client credentials, and use http://localhost:8080/callback as the redirect URI. The repository provides an .env.example file for local configuration. The documented MCP client configuration passes KOMMO_SUBDOMAIN and KOMMO_ACCESS_TOKEN in the server environment.

You can install the Kommo CRM MCP server with uvx --from kommo-kiro-power kommo-mcp, install the package from source with pip install -e ., or install the released MCPB bundle in Claude Desktop or another MCPB-compatible client. Kiro users can import the GitHub repository as a custom Power. After installation, run python scripts/oauth_setup.py to authorize the account, then start the server.

Tools and capabilities

The Kommo CRM MCP server groups its tools by CRM object:

• Leads: list, retrieve, create, update, delete, bulk update, and move between stages.
• Contacts: list, retrieve, create, and update.
• Pipelines and stages: list, create, and update.
• Tags: list, add, and remove; missing tags are created when needed.
• Tasks: create and list, including an overdue filter.
• Notes: add a note to a lead.
• Chat: list message templates and send a message.
• Companies: create and list.
• Custom fields: list and create several supported field types.
• Combined workflow: create a lead, contact, and company in one call.

Limitations and notes

The server requires a Kommo account, an OAuth integration, and valid credentials. It is designed for Kommo API v4 and requires Python 3.10 or newer. The documented transport is stdio, so the client must support launching a local MCP process or installing the provided MCPB bundle.

The server runs on the user’s machine and communicates with that user’s Kommo account API. The README states that it has no telemetry and does not send data to the author or other third parties. Rate limiting and caching affect request timing: no more than five requests per second are allowed, and cached pipeline, stage, and field data can remain for up to 10 minutes.