monday.com MCP

The log-activity skill allows you to manage the activity feed (Timeline) of CRM items like deals, contacts, leads, or accounts. You can use it to log new activities (calls, meetings, notes), read activity history, or update existing timeline entries.

Common User Triggers

• Log: "log my call with Acme", "add a note to the deal", "record this meeting", "I just spoke with..."
• Read: "what activities happened on TechCorp this week", "show timeline for Globex", "activity history for..."
• Update: "update the note on the Acme deal", "edit the meeting note"

Core Concepts

• Timeline: The activity feed on any CRM item. It includes calls, meetings, notes, and emails.
• Structured vs. Notes: The skill attempts to create structured activities (e.g., a 'Call' with specific attendees and outcomes). If structured logging is unavailable or fails, it gracefully degrades to a free-text note using create_timeline_note.
• HITL (Human-In-The-Loop): All write and update operations require user confirmation before execution.
• Append-Only: The skill supports creating and updating, but does not support deleting timeline entries.

"[activity description — 'call with Acme about renewal', 'note on Globex deal: demo next week', 'activities on TechCorp this week']"

The run-sequence skill allows you to manage monday CRM sequences end-to-end via natural language. You can list sequences, create new ones, enroll contacts, activate/deactivate flows, duplicate sequences, and track performance analytics.

Common user requests include:

• "What sequences do I have?"
• "Create a welcome sequence"
• "Enroll Acme in the welcome sequence"
• "Show me sequence analytics for Q2 nurture"
• "Deactivate the cold outreach sequence"
• "Who is enrolled in the nurture sequence?"

Key Constraints & Behaviors:

• No Editing: You cannot add, modify, or remove steps in a sequence via MCP. For editing, users must go to Tools → Sequences in the monday CRM UI.
• Deactivation vs. Pausing: Deactivating a sequence (deactivate-sequence) does not stop existing enrollments; it only blocks new enrollments from starting.
• Sequence States:
  • ACTIVE: New enrollments are allowed and running.
  • INACTIVE: New enrollments are blocked; existing runs continue.
• Board Resolution: Most operations (list, enroll, contact-journey) require a board_id. You must resolve the board name to a board_id before calling the specific sequence tools.

The skill translates user descriptions into internal archetypes to propose a board structure. The user never sees these archetype labels; they only see the resulting board proposal using their own terminology.

Custom Mode

If no archetype fits, the skill composes a structure from:

• Primary entity board: Named using user's language (e.g., Candidates, Listings, Matters).
• Pipeline mechanism: Groups (low-volume/high-touch) or Status column (high-volume).
• Secondary board: If two distinct entities are mentioned.
• Columns: Standard set (Owner, Stage, Value, Probability, Date, Source, Notes) plus 1–3 industry-specific columns.

When using the data-cleanup skill, the following safety and execution patterns are enforced to protect your data:

Safety Rails

• No Deletes: The skill will never delete items or columns.
• No Amount-Column Writes: To maintain forecast integrity, the skill does not automatically write to amount columns.
• No Cross-Workspace Moves: Items are never moved between workspaces.
• Flag-only items: Items that require human judgment (like missing amounts) are flagged in the summary but are never written to automatically.

Execution Constraints

• Bounded Writes: Execution is capped (e.g., at 500 writes) to prevent runaway automation.
• Concurrency Control: Writes are performed in controlled batches (e.g., 5 concurrent operations) to respect API limits.
• Auditability: Every run generates a document containing a before/after log. These docs include a <!-- claude-skill-id: data-cleanup --> comment and a Generated by Claude footer to ensure idempotency and traceability.

When running the data-cleanup skill, the following logic is applied to specific data types:

• Phone format → E.164: Strips non-digit/+ characters. If no leading + exists, it prepends the dial code based on the populated country column or known prefixes (e.g., 1 for US). Skips if length is <7 or >15 digits.
• Email lowercase: Converts values to lowercase using value.toLowerCase(). Only writes if the value actually changes.
• Country code from phone: Parses the leading dial code to an ISO-2 country code and writes to the country column using the all_monday_api shape: {"countryCode": "<ISO2>", "countryName": "<name>"}.
• Country code from email CCTLD: Infers country if the email ends in a specific TLD (e.g., .uk, .de) and the country column is empty.
• Email guess: Uses <firstname>.<lastname>@<companydomain>. Warning: This is a heuristic and may be inaccurate. Every guessed email is tagged with a Source = guessed status flag.
• Bulk-set (Owner/Source/Stage): Applies values within a user-confirmed scope (group or status filter) using change_item_column_values.
• Backfill stale last-touch: Sets the date to today only if the cell is empty or older than 90 days. It will never overwrite a more recent date.

The meeting-to-deal skill includes a safety mechanism for handling ambiguous matches. An ambiguity occurs when a meeting's attendees or domain match multiple deals with near-equal scores.

Ambiguity Criteria:

• The top match score and the second-best match score are equal (or the gap between them is less than 3).

Resolution Workflow:

1. Detection: Instead of guessing or skipping, the agent flags the ambiguity.
2. User Prompt: The agent presents the candidate deals to the user and asks for clarification.
3. Natural Language Selection: The user can resolve the ambiguity using natural language (e.g., "first to Series B, second to Series C" or "both to Series B").
4. Resumption: The skill resumes from the matching step without re-running the initial discovery steps (user context, meeting retrieval, or board resolution).
5. Audit Trail: Resolved ambiguous matches are logged in the generated sync document under a Matched (X/X) — user-confirmed section for auditing purposes.

The meeting-to-deal skill operates in three modes to control how much automation is applied to your monday.com boards.

Modes

Safety Rails

Regardless of the mode selected, the skill enforces the following restrictions to prevent data loss or corruption:

• No Deletes: The skill will never delete items or updates.
• No Amount-Column Writes: The skill is strictly forbidden from modifying any column related to monetary amounts.
• No Cross-Workspace Moves: Items will not be moved between different workspaces.
• Batched Stage Transitions: Stage transitions are always presented as a batched plan for review; they are never written one-off or silently.