Kylas CRM MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Kylas CRM MCP Server Find leads that have been idle for more than 7 days"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Kylas CRM MCP Server (Lead Only)
Model Context Protocol server for Kylas CRM lead operations. Use it from Cursor, Claude Desktop, or any MCP client to create leads, search and filter leads, and look up users, products, and pipelines.
Features
get_lead_field_instructions – Get lead schema (standard + custom fields, picklist IDs)
create_lead – Create a lead with dynamic fields from user context
search_leads – Search/filter leads by multiple criteria
lookup_users – Resolve user names to IDs (for owner, created by, etc.)
lookup_products – Resolve product names to IDs
lookup_pipelines / get_pipeline_stages – Resolve pipeline and stage for open/closed/won leads
search_idle_leads – Find leads with no activity for N days
Related MCP server: Kylas CRM MCP Server
Requirements
Python 3.10+
Kylas account and API key
Installation
pip install -e .
# or from PyPI (after publish): pip install kylas-crm-mcpConfiguration
Set environment variables (or use a .env file):
Variable | Required | Description |
| Yes | Your Kylas API key |
| No | API base URL (default: https://api.kylas.io/v1) |
Running the server
The server uses stdio transport (default for MCP). Run:
python -m kylas_crm_mcp
# or: python main.py (when developing from repo root)MCP clients (e.g. Cursor) typically start this process and communicate via stdin/stdout.
Docker
docker build -t kylas-crm-mcp .
docker run -e KYLAS_API_KEY=your_key -i kylas-crm-mcpDevelopment
pip install -e ".[dev]"
pytestLicense
See repository for license information.
Available Tools
8 toolscreate_leadA
Create a lead in Kylas CRM with only the fields the user wants (no static field list).
You MUST call get_lead_field_instructions FIRST to get valid API names and Field IDs. Infer from user context which fields to send; include only those in field_values.
field_values: Map of field identifier to value.
Standard fields: use API name as key at top level (e.g. firstName, lastName, companyName, emails, phoneNumbers, leadSource, isNew).
Custom fields: MUST be under "customFieldValues" with internal name as key (e.g. "customFieldValues": {"cfLeadCheck": "Checked"}). Do not use field ID as key—Kylas expects internal names. If you pass a field ID (e.g. "1210985"), the server will resolve it to the internal name (e.g. cfLeadCheck) automatically.
For a single email use "email": "user@example.com"; for phones "phone": "5551234567" and optionally "phone_country_code": "+1".
For picklists use the Option ID (number) from the cheat sheet.
Inputs: field_values (required). Returns: result
get_lead_field_instructionsA
Get all lead fields for the current tenant. CALL THIS FIRST before creating a lead. Returns a cheat sheet with API names (standard fields), Field IDs (custom fields), and Picklist Option IDs. Use this to build field_values for create_lead based on what the user wants—do not use static fields.
Inputs: no named parameters. Returns: result
get_pipeline_stagesA
Get stages for a pipeline. Call this only after the user has confirmed which pipeline to use (from lookup_pipelines). Do not call before pipeline confirmation. Returns pipeline name and list of stages for that pipeline only, with id, name, and forecastingType (OPEN, CLOSED_WON, CLOSED_LOST, CLOSED_UNQUALIFIED). Use the stage IDs in search_leads: filters [{"field": "pipeline", "operator": "equal", "value": pipeline_id}, {"field": "pipelineStage", "operator": "equal", "value": stage_id}]. If the user said "open leads" or "closed leads" and more than one stage has the same forecastingType, ask which stage they mean. pipeline_id: The pipeline ID (from lookup_pipelines).
Inputs: pipeline_id (required). Returns: result
lookup_pipelinesA
Look up pipelines by name (for leads). Use when the user asks for leads by stage (e.g. open/closed/won/lost) but does not specify which pipeline.
Call this first; do NOT call get_pipeline_stages until after the user confirms the pipeline.
Present the pipeline(s) (id and name) and ask the user which pipeline they mean. If only one pipeline is found, still ask for confirmation.
Only after the user confirms, call get_pipeline_stages with that pipeline ID to get stages for that pipeline, then search_leads. query: Search string. Use "name:<pipeline_name>" or just the pipeline name; empty string returns all pipelines for the entity. entity_type: Entity type (default LEAD). page: 0-based page (default 0). size: Max 50 (default 50).
Inputs: query, entity_type, page, size. Returns: result
lookup_productsA
Look up products by name. Use this BEFORE filtering leads by product when the user gives a product name.
If one product is found, use that product's ID in search_leads (e.g. {"field": "products", "operator": "equal", "value": }).
If multiple products are found, ask the user which product they mean (list the options), then use the chosen product's ID in search_leads. query: Search string. Use "name:<product_name>" (e.g. "name:Widget") or just the product name (e.g. "Widget"); the server will send name:value to the API. page: 0-based page (default 0). size: Max 50 (default 50).
Inputs: query (required), page, size. Returns: result
lookup_usersA
Look up users by name, or list all users in the system.
Use return_all=True (with query "name:" or empty) to fetch all users in one response (all pages combined).
For name search: query in field:value form (e.g. "firstName:last", "name:Last"). If one user is found, use that ID in search_leads; if multiple, ask which one. query: Search string (e.g. "firstName:last", "name:Last"). Use "name:" or leave default to list all when return_all=True. page: 0-based page (default 0). Ignored when return_all=True. size: Page size, max 50 (default 50). Used per page when return_all=True. return_all: If True, fetch all pages and return every user in one response (cap 500).
Inputs: query, page, size, return_all. Returns: result
search_idle_leadsA
Search for idle/stagnant leads: no activity for at least the given number of days. Uses both updatedAt and latestActivityCreatedAt; a lead is returned only when BOTH dates are on or before (today − days), so the effective last activity is before the threshold.
days: Minimum days with no activity (e.g. 10 for "no activity since 10 days"). time_zone: IANA timezone for threshold (e.g. America/New_York). Default: Asia/Calcutta. page: 0-based page (default 0). size: Page size, max 100 (default 20). sort: Sort e.g. "createdAt,desc" (default).
Inputs: days (required), time_zone, page, size, +1 more. Returns: result
search_leadsA
Search/filter leads. Only fields marked [FILTERABLE] in get_lead_field_instructions can be used. Call get_lead_field_instructions first to get filterable fields and their types.
filters: List of filter objects. Each must have:
field (str): Field internal/API name (e.g. firstName, country, source, createdAt).
operator (str): One of the allowed operators for that field type (e.g. equal, contains, greater).
value: Value to compare. For PICK_LIST/MULTI_PICKLIST use Option ID (number), except requirementCurrency, companyBusinessType, country, timezone, companyIndustry — use internal name (string). For date/datetime (incl. custom e.g. cfDateField): value null for today/is_null/is_not_null; single ISO string for greater/greater_or_equal/less/less_or_equal e.g. "2026-02-02T18:30:00.000Z"; for between use [startISO, endISO].
timeZone (str, optional): For date/datetime filters only; default from server or env.
type (str, optional): Field type from cheat sheet. If omitted, inferred from schema. For user look-up fields (createdBy, updatedBy, convertedBy, ownerId, importedBy): value must be user ID (number). Call lookup_users first. For the products field: value must be product ID (number). Call lookup_products first; if multiple matches, ask which product, then use that ID here. For pipeline / pipelineStage (e.g. open leads, closed leads): call lookup_pipelines first, ask the user to confirm which pipeline, then call get_pipeline_stages for that pipeline only; if stage is ambiguous ask which stage, then use pipeline + pipelineStage filters here. page: 0-based page (default 0). size: Page size, max 100 (default 20). sort: Sort e.g. "createdAt,desc" (default).
Operators by type (examples): TEXT_FIELD: equal, contains, is_empty. NUMBER: equal, greater, between, is_null. PICK_LIST: equal, in, is_null. DATETIME_PICKER: today, yesterday, between, is_not_null, greater, less, current_week, etc.
Inputs: filters (required), page, size, sort. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 8 tools
Most tools have distinct purposes: create_lead, get_lead_field_instructions, search_leads, and search_idle_leads are clearly differentiated. However, the four lookup tools (lookup_pipelines, lookup_products, lookup_users, get_pipeline_stages) could cause confusion as they all serve similar 'lookup' functions but for different entities, requiring careful reading of descriptions to distinguish them.
The naming follows a consistent verb_noun pattern (e.g., create_lead, lookup_pipelines, search_leads) with snake_case throughout. The only minor deviation is get_lead_field_instructions, which uses 'get' instead of 'lookup' or 'search', but it still fits the verb_noun style and is understandable.
With 8 tools, the count is well-scoped for a CRM server, covering core operations like lead creation, field instructions, searching, and lookups for pipelines, products, and users. Each tool serves a specific purpose without bloat, aligning with typical CRM workflows.
The toolset provides strong coverage for lead management, including creation, searching (with filters and idle leads), and necessary lookups. A minor gap is the lack of update or delete operations for leads, which might limit full CRUD lifecycle management, but agents can still handle core tasks effectively with the available tools.
Maintenance
Related MCP Connectors
Read noCRM.io leads, pipeline steps, comments and client folders; create, update and assign leads.
Search, summarize, and update your RevOrbit CRM (leads, deals, contacts, tasks) in plain language.
I manage your Support Agent CRM, pipeline analytics, follow-ups, Leads Discovery, and workspace out…
List leads, contacts, companies and tasks; create and update leads in Kommo.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to perform secure Salesforce Lead management operations including CRUD actions, status updates, and idempotent syncing. It features TRD-compliant audit logging and OAuth 2.0 authentication to ensure reliable CRM interactions.-

Kylas CRM MCP Serverofficial
AlicenseAqualityDmaintenanceModel Context Protocol server for Kylas CRM lead operations, enabling lead creation, search/filter, and lookup of users, products, and pipelines.1412MIT- FlicenseNot gradedqualityDmaintenanceManages leads and sends emails directly from Claude chat, with tools to add, search, filter, and update leads, plus email capabilities.-
- FlicenseNot gradedqualityDmaintenanceEnables ChatGPT to interact with noCRM leads for outreach tracking, including listing, retrieving, creating leads and adding comments, without destructive operations.-