LACRM MCP Server

A Model Context Protocol (MCP) server for Less Annoying CRM that provides comprehensive API access through 87 tools.

Published as lacrm-mcp on npm.

Current version: 1.3.1

Key Features

• 87 tools covering contacts, pipelines, tasks, events, notes, emails, files, relationships, groups, bulk CSV, and settings

• Name resolution: Search tools accept human-readable names (status names, user names, calendar names, custom field names) and auto-resolve to IDs at runtime — no prerequisite lookup calls needed

• count_only mode: All search/list tools support count_only: true which auto-paginates through all results and returns accurate totals with categorical breakdowns — no manual pagination required

• Flat-string shortcuts: email_address, phone_number, website_url on contact create/edit auto-convert to the required array format

• Response summaries: List-returning tools wrap results in {summary, results} envelopes with machine-counted page counts and breakdowns

• ID sanitization: Defense-in-depth stripping of accidental quote characters from ID parameters

• Rate limiting: Client-side enforcement of 120 requests/minute with automatic waiting

Related MCP server: pipedrive-mcp

Installation

From npm (recommended for production)

npm install -g lacrm-mcp

From source

git clone https://github.com/optimize-overseas/lacrm-mcp.git
cd lacrm-mcp
npm install
npm run build

Configuration

Environment Variable (Recommended)

Set the LACRM_API_KEY environment variable:

export LACRM_API_KEY=your-api-key

Config File

Alternatively, create a config file at ~/.lacrm-config.json:

{
  "apiKey": "your-api-key"
}

Claude Desktop Configuration

Add to your Claude Desktop configuration file (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "lacrm": {
      "command": "node",
      "args": ["/path/to/lacrm-mcp/build/index.js"],
      "env": {
        "LACRM_API_KEY": "your-api-key"
      }
    }
  }
}

Testing with MCP Inspector

npm run inspector

MCP Resources

This server provides workflow resources that help AI clients understand how to use the MCP effectively. Resources are discoverable via the standard MCP resources/list protocol.

Usage: Call resources/list to discover available resources, then resources/read with the URI to get the content.

Response Format

Summary Envelopes

All list-returning tools wrap their results in a structured {summary, results} envelope. The summary provides machine-counted page counts and categorical breakdowns, which prevents LLMs from miscounting items in large JSON arrays -- a known failure mode when datasets exceed approximately 50 items.

Envelope structure:

{
  "summary": {
    "page_count": 15,
    "has_more_results": false,
    "by_status": {
      "Active": 10,
      "Closed - Won": 3,
      "Closed - Lost": 2
    }
  },
  "results": [
    { "...full pipeline item data..." },
    { "...full pipeline item data..." }
  ]
}

The summary object always includes:

• page_count -- the number of items on this page only (not the total across all pages)

• has_more_results -- whether the API has additional pages beyond this response

• note (when has_more_results is true) -- a reminder that page_count is partial and count_only=true should be used for accurate totals

Important: When has_more_results is true, the page_count value is NOT the total number of matching results. Use count_only=true for accurate totals across all pages.

Depending on the tool, the summary may also include one or more categorical breakdowns that group items by a relevant field.

Tools with Summary Envelopes

Single-record tools (e.g., get_contact, get_task, get_pipeline_item) return the raw API response directly without a summary envelope.

Name Resolution (v1.3.0)

Many LACRM API parameters require internal IDs (status IDs, user IDs, calendar IDs, custom field IDs). In v1.3.0, name-based alternatives were added so the caller can pass human-readable names instead. The server resolves names to the values the LACRM API expects at runtime — status/user/calendar IDs, and (for custom fields) the field name LACRM writes by. All lookups are case-insensitive, and if a name cannot be resolved the error message lists all available options.

These parameters are instance-agnostic -- they work with any LACRM account regardless of how statuses, users, calendars, or custom fields are configured.

Name Resolution Parameters

How It Works

1. The caller passes a name-based parameter (e.g., status_name: "Active")

2. The server queries the relevant API endpoint (e.g., GetPipelineStatuses)

3. A case-insensitive match resolves the name to its ID

4. The resolved ID is used for the actual API call

5. If no match is found, an error lists all available names

For user_name_filter, matching works against full name ("First Last"), first name only, or last name only.

For custom_field_names, dropdown fields are additionally validated -- if the supplied value is not among the allowed options, an error lists the valid choices.

Example

Instead of:

{
  "pipeline_id": "abc123",
  "status_id": "def456",
  "custom_fields": { "cf_789": "Matt" }
}

Use:

{
  "pipeline_id": "abc123",
  "status_name": "Active",
  "custom_field_names": { "Hunter": "Matt" }
}

Count Mode (v1.3.0)

All search and list tools support a count_only parameter. When set to true, the tool auto-paginates through all pages (up to a safety cap of 100 pages) and returns only aggregate counts with breakdowns -- no results array is included. This is useful for questions like "how many deals are in each status?" without transferring the full dataset.

Supported Tools

Count Mode Response

{
  "total": 47,
  "breakdowns": {
    "by_status": {
      "Active": 30,
      "Closed - Won": 12,
      "Closed - Lost": 5
    }
  }
}

When count_only is true:

• The tool paginates through all available pages automatically

• No results array is returned -- only total and breakdowns

• A safety cap of 100 pages prevents runaway pagination

Flat-String Shortcuts (v1.3.0)

create_contact and edit_contact accept simplified string parameters for common single-value fields:

Each shortcut is mutually exclusive with its corresponding array parameter (e.g., email_address cannot be used together with email). This eliminates the need to construct array-of-objects structures for the common case of a single value.

Example

Instead of:

{
  "name": "Jane Doe",
  "email": [{"Text": "jane@example.com", "Type": "Work"}],
  "phone": [{"Text": "555-0100", "Type": "Work"}]
}

Use:

{
  "name": "Jane Doe",
  "email_address": "jane@example.com",
  "phone_number": "555-0100"
}

Bulk CSV Operations (v1.4.0)

Import or update contacts in bulk from a CSV, paced to LACRM's 1-request/second guidance. Four tools cover the workflow, and everything is instance-agnostic: the caller supplies the entire field configuration (which CSV columns map to which LACRM fields, how each one merges, and any create-time defaults) as tool arguments. No field names, merge rules, or use cases are built into the server.

Template column order: generated templates list LACRM built-in fields first — the key column, then the owner name, then the address block, then any other standard fields (Email, Phone, etc.) — and custom fields last. Columns read left-to-right as identity → standard → custom, regardless of the order fields are supplied in.

Update merge model (column-presence)

Bulk updates are read-merge-write per contact, so a partial CSV never clobbers fields it does not mention. A column absent from the CSV is always left unchanged. When a column is present, its strategy decides what happens:

Use replace for fields that should mirror the CSV exactly (a blank cell means "clear it"); preserve_if_blank for fill-only fields; union_semicolon for accumulate-style list fields; never_write to lock a field against this operation.

Update address columns (append-if-absent, v1.5.0)

Update mode can also carry a structured address via address_config (same shape as create's address: street1, optional street2/city/state/zip/country, and a literal type defaulting to Work). Its mapped CSV columns are added to the generated template and recognized by validation. For each row the uploaded address is appended to the contact only if it is not already on the record — compared case- and format-insensitively on Street/City/State/Zip, with street-suffix and directional abbreviations canonicalized (St↔Street, N↔North), a US ZIP+4 reduced to its 5-digit base (international postal codes are kept whole, so two codes that differ only past the 5th character stay distinct), and a full state name matched to its 2-letter code. On a match the existing CRM address is kept and the file's value is ignored; existing addresses are never modified, reordered, or removed — the new one is appended at the end (the primary/mailing address at position 0 is untouched). Because LACRM's EditContact replaces the entire address array, the worker reads the contact's existing addresses and writes back the complete merged array. One address per row.

Throttle & detached execution

bulk_execute spawns a separate worker process that calls LACRM strictly sequentially at >=1s spacing, so a multi-thousand-row run (which can take hours) proceeds independently of the request that started it and survives MCP/host restarts. Run state is persisted to disk — directory configurable via LACRM_BULK_RUNS_DIR (default: a lacrm-bulk-runs folder under the OS temp dir). A crashed run resumes from where it left off if re-launched against the same run's spec.

Typical workflow

1. bulk_generate_template -> hand the user the CSV to fill and the per-field behavior report.

2. bulk_validate_csv -> review row counts, which fields will change vs. be preserved, and any blocking errors. Always preview before executing.

3. bulk_execute with confirm: true -> receive a run_id.

4. bulk_run_status -> poll until status is completed, then fetch the report CSV.

Available Tools

Discovery Tools (10)

Contact Tools (6)

Bulk CSV Tools (4)

See Bulk CSV Operations for the full workflow and merge model.

Event Tools (6)

Task Tools (6)

Note Tools (6)

Pipeline Item Tools (7)

Email Tools (5)

File Tools (3)

Relationship Tools (5)

Group Membership Tools (4)

Custom Field Settings Tools (4)

Group Settings Tools (4)

Pipeline Settings Tools (4)

Pipeline Status Settings Tools (4)

Team Settings Tools (5)

Webhook Settings Tools (4)

Custom Fields

Using custom_field_names (Recommended -- v1.3.0)

Pass field names directly as keys. The server resolves names to IDs automatically and validates dropdown values:

{
  "custom_field_names": {
    "Lead Source": "Referral",
    "Deal Value": 50000,
    "Expected Close": "2025-03-15"
  }
}

Invalid field names return an error listing available fields. Invalid dropdown values return an error listing valid options.

Using custom_fields (verbatim, by field name — no validation)

LACRM v2 writes custom fields by their name at the top level; a numeric CustomFieldId is silently ignored on write. custom_fields writes its keys verbatim, so use field names as keys. Use custom_field_names instead when you want dropdown/option validation.

1. Call get_custom_fields with record_type="Contact" or record_type="Company" (for contacts/companies) or get_pipeline_custom_fields with the pipeline_id (for pipeline items)

2. Note the field names, types, required status, and valid options

3. Use field names as keys in the custom_fields parameter

Example:

{
  "custom_fields": {
    "Hunter": "Matt",
    "Deal Value": 50000,
    "Expected Close": "2025-03-15"
  }
}

Pipeline Support

Workflow for creating pipeline items:

1. Call get_pipelines to discover pipeline IDs

2. Use create_pipeline_item with pipeline_id and either status_name (auto-resolves, v1.3.0) or status_id

3. Optionally include custom_field_names (v1.3.0) or custom_fields for pipeline custom fields

The tools provide clear error messages when required fields are missing or names don't match.

API Reference

This MCP server wraps the Less Annoying CRM API v2. All API calls are made to https://api.lessannoyingcrm.com/v2/.

Rate Limiting

The MCP server enforces rate limiting to protect LACRM's API:

• Limit: 120 requests per minute (sliding window)

• Behavior: If the limit is reached, requests automatically wait until a slot is available

• No configuration needed: Rate limiting is always active

This ensures that even aggressive AI usage won't overwhelm LACRM's servers.

Security

Authentication

• API key is loaded from LACRM_API_KEY environment variable (recommended)

• Fallback to config file at ~/.lacrm-config.json

• API key is never logged or exposed in error messages

ID Parameter Sanitization

The server includes defense-in-depth sanitization for all ID parameters. LLMs sometimes over-quote values, passing "\"86441\"" (a string with embedded literal quotes) instead of the clean string "86441". The LACRM API rejects values with embedded quotes as invalid UIDs.

To guard against this, the client automatically strips surrounding single and double quote characters from any parameter whose key ends with:

• *Id -- single ID parameters (e.g., ContactId, PipelineId)

• *Ids -- array-of-ID parameters (e.g., ContactIds)

• *IdList -- ID list parameters

When sanitization occurs, a warning is logged so the issue can be traced. Other string parameters (names, notes, descriptions, etc.) are never modified.

File Upload Security

The create_file tool validates file paths to prevent:

• Path traversal attacks (../ sequences)

• Access to sensitive directories (.git, .env, .ssh, .aws, node_modules, etc.)

Input Validation

• All tool inputs are validated with Zod schemas

• Error responses use isError: true flag for LLM-friendly handling

• No shell command execution - all operations use direct API calls

License

MIT