MCP recipes

A daily reply digest in Claude: who answered, who needs you

This recipe asks your assistant to find the leads who replied most recently, read the last messages and the pending approvals, and write a short digest: who answered, what they want and the three conversations to handle first. It only reads.

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

What this recipe does

A three step read followed by a written summary. It treats the last message of each thread as the thing to act on and it separates replies that need you from replies the agent can handle.

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.

The prompt

Paste into Claude or ChatGPT

Write my daily reply digest for Sellinger.
1. Call list_leads with status RESPONDED, ordered by updated_at descending, per_page 20.
2. For the first eight, call get_lead_conversation with limit 5.
3. Call list_approvals with status pending.
Then write: Replies that need me (name, company, what they asked, the last message), Replies the agent can handle, and Approvals waiting. End with the three conversations I should open first. Do not send or approve anything.

The exact tools it calls

Tools used by this recipe
Tool	What the recipe uses it for
`list_leads`	Leads in RESPONDED, newest first
`get_lead_conversation`	The last messages of each thread, newest first
`list_approvals`	Drafts waiting on a decision

The connector has 43 tools in total.

What needs your approval

Nothing. Every tool in this recipe reads.

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 change on the lead record, not strictly the last message, so ordering is approximate.
Auto replies and out of office messages can appear as replies. The assistant reads the text, so ask it to ignore them.
Run it from the same saved prompt each morning. Scheduling depends on your assistant.

Common questions

Does it mark replies as read?

No. It only reads.

Why RESPONDED?

It is the lead status for people who answered. Pipeline stages such as ENGAGED are filtered with the stage parameter, not status.

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.