MCP recipes

Weekly pipeline review from Claude: wins, risks and three actions

This recipe asks your assistant to read your pipeline and write a short review: wins, risks and stuck leads, plus exactly three recommended actions. It only reads. Nothing is sent or changed unless you then ask for it.

Published by Sellinger AI. Numbers and settings checked on 19 October 2026 against the fact register. Report an error.

What this recipe does

A five step read of the workspace followed by a written summary. It uses an updated at sort to find the least recently touched leads and confirms with their conversations before calling them stuck.

Connect the Sellinger connector first (claude.ai: Settings, Connectors, Add custom connector, paste https://app.sellinger.ai/api/v2/mcp). The steps are on the MCP page.

In claude.ai this is also available as the saved prompt weekly_pipeline_review, so you can start it from the prompt menu without typing.

The prompt

Paste into Claude or ChatGPT

Run a weekly pipeline review of my Sellinger workspace.
1. Call get_pipeline_counts.
2. Call get_analytics with period 30d.
3. Call list_meetings for the next 7 days with status scheduled.
4. Call list_approvals with status pending.
5. Call list_leads for stages ENGAGED and OUTREACH, ordered by updated_at ascending, and check a few of the oldest with get_lead_conversation.
Then write three sections: Wins, Risks, Stuck leads (name, stage, days since updated). Finish with exactly 3 recommended actions, each naming the lead, campaign or approval and the tool that would do it. Do not act yet.

The exact tools it calls

Tools used by this recipe
Tool	What the recipe uses it for
`get_pipeline_counts`	Lead counts per pipeline stage
`get_analytics`	Total leads, conversion rate, meetings booked and response rate, each against the previous window
`list_meetings`	Meetings booked in the next 7 days
`list_approvals`	Items waiting on a person
`list_leads`	Least recently touched leads in ENGAGED and OUTREACH
`get_lead_conversation`	Confirms a lead is truly stalled and not just recently edited

The connector has 43 tools in total.

What needs your approval

Nothing. Every tool in this recipe reads.
Recommended actions that would change data, such as update_lead, approve_approval or update_campaign, wait until you say yes.

Sample output

No sample output is published for this recipe yet. It has not been run end to end in Claude or ChatGPT for this page, and we do not invent output. The tool names and what each returns are taken from the connector's tool descriptions in the product code (checked 19 October 2026). When a real run is captured it will be added here with its date.

Notes and limits

Updated at is the last time anything changed on the lead record, not strictly the last message, so the recipe treats it as an approximation.
The status filter on meetings takes one value, so the saved prompt asks for scheduled meetings and drops completed, cancelled and no show rows.

Common questions

Does this change anything in my CRM?

No. All six tools read. Actions it recommends wait for your confirmation.

Can I schedule it weekly?

Run it from the saved prompt each week. Scheduling depends on your assistant, not on Sellinger.

Which assistants work?

Claude, ChatGPT and other clients that support the streamable HTTP transport with sign in.

Sources and method

Tool names, parameters and warnings are from frontend/src/lib/mcp/hosted-tools.ts and the saved prompts in hosted-prompts.ts (43 registered tools, checked 19 October 2026). The recipe has not been run for this page.