Skip to main content

Overview

Ask in one sentence which LP prospects of a given type have gone quiet past a certain point, and who originally introduced you to them. The result is a specific, warm ask you can hand to a connector before you meet them. Done in the app, this is a saved view you would have to build or several separate searches you would have to reconcile in your head. Examples below come from a live demo environment. Company and contact names are fictitious.

What you can do

  • Find prospects on an LP list that your firm has not spoken to in a defined window, based on the most recent logged meeting or call
  • Narrow that by organization type, for example family offices, endowments, or funds of funds
  • Filter by who originally made the introduction, so you can hand each connector a list of only their own dormant introductions
  • Write the outcome back to the list so the follow-up is tracked rather than living in your head

Prerequisites

  • Affinity MCP connected in Claude (or another MCP client)
  • An LP prospects list with an organization type field populated
  • A field or note capturing who made the original introduction, since Claude filters on data that exists
  • Email and calendar sync turned on, so “when we last spoke” reflects reality rather than what people remembered to log
The silence window is only as accurate as your logged interactions. If meetings and calls are syncing, this works on its own. If your team logs sporadically, expect false positives on the quiet list.

Use case 1: Find the prospects who have gone quiet

When to use it: You want to know where your LP pipeline has gone cold, without building a view for it. How to do it: Ask Claude with the Affinity MCP connected: “Find [organization type] on our [LP prospects list name] that we haven’t spoken to in [90] days, based on the most recent logged meeting or call.” Several filters are stacked into one sentence here: the list, the organization type, and the recency test anchored to actual logged interactions rather than to record creation or last edit. In the demo this returned in seconds, against a task that otherwise means building a saved view or reconciling separate searches manually. Tip: Say “based on the most recent logged meeting or call.” Without that anchor, “haven’t spoken to” is ambiguous and the results will be less trustworthy.

Use case 2: Filter by who made the introduction

When to use it: You are about to meet a connector who has been a good source of introductions, and you want to hand them something specific to work on. How to do it: Ask: “Find [organization type] on our [LP prospects list name] that we haven’t spoken to in [90] days, based on the most recent logged meeting or call, and were originally introduced by [name].” This is the filter that turns a stale-pipeline report into a usable ask, and in the demo it was stacked into the same single sentence as the recency test rather than asked as a follow-up. Four family offices came back, all originally introduced by the person the partner was about to meet, all quiet for more than 90 days. That is a concrete, warm request rather than a general “please keep us in mind.” Note: This only works if the introduction source is captured somewhere Claude can read, whether that is a dedicated field or a note on the record. If it is not being captured today, that is the prerequisite worth fixing first.

Use case 3: Log the outcome back to the list

When to use it: You have made the ask and you want the follow-up tracked, without opening each record. How to do it: Ask: “On [LP prospects list name], log a note on each of those records saying I asked [name] to re-introduce us on [date], and set the follow-up status to [status].” This writes back to the same records you just queried, in the same chat. Logging the ask at the moment you make it is what keeps this from becoming a search you re-run in three months with no memory of what you already did. Because the MCP writes as well as reads, the whole loop stays in one conversation.

Tips for getting better results

  • Ask the way you would ask a colleague. The demo prompts were full sentences with the filters stated plainly, not filter syntax.
  • Anchor the recency test to logged meetings or calls explicitly, so the answer is not based on when the record was last touched.
  • Stack the filters into one sentence rather than asking three narrowing questions in sequence. In the demo, the list, the organization type, the recency test, and the introduction source were all said in a single sentence and handled in one pass.
  • Run this before the meeting with the connector, not after. The output is only valuable if you can hand it to someone in the room.
  • If the introduction source is not being captured consistently, fix that first. Every filter in this recipe depends on data your team has actually logged.

Troubleshooting & FAQ

A prospect I know we spoke to recently showed up as quiet. Check that the person who spoke to them has email and calendar sync on and has been logging in regularly, since sync can switch off for users who stay logged out for long enough. Consistent note-taking on calls closes the rest of the gap, and the MCP itself can write those notes back for you. Can the MCP see records on lists I do not have access to? No. You authenticate as yourself and the MCP respects your exact CRM permission set, including list visibility and note visibility. How do I keep the underlying data current enough for this to be trustworthy? Three things. Make sure your team is logging in regularly so sync stays on, keep good practice around recording notes and maintaining lists, and use the MCP itself to write data back, which is often the lowest-friction way to keep records current. If you want a structured data audit for your instance, your customer success manager can run a workshop on it. What if I write something wrong back to the list? As of this session the MCP does not have a delete function. For anything touching multiple records, ask Claude to show you the proposed changes before committing them. Affinity’s change log will show what was changed, when, and who changed it. Which tiers include MCP access? Scale, Advanced, and Enterprise, at no additional cost. You need your own account with Claude, ChatGPT, or Copilot, and those subscription and token costs sit between you and that provider.