Summarize the WhatsApp Groups You Never Read with Claude: A Daily Digest Prompt (2026)

You are in a dozen WhatsApp groups. Clients, the team, the building committee, the school parents. You muted most of them, and somewhere in there is the one message that asked you something on Tuesday.

This guide shows how to use a WhatsApp MCP connector in Claude to turn those groups into one short digest. (New to the connector itself? Start with the hosted WhatsApp MCP service, or compare options in the best WhatsApp MCP servers.) One prompt, about three read-only tool calls, and nothing gets sent.

What can a WhatsApp MCP do with the groups you never read?

With the Blueticks WhatsApp MCP server connected, one prompt in Claude reads the last few days of your unread group chats. It returns the decisions made, messages that name you or ask you something, and any dates or deadlines. The digest uses read actions only. Nothing is sent.

The Model Context Protocol is, in its own words, "an open-source standard for connecting AI applications to external systems." The Blueticks server is one such connection. It exposes your linked WhatsApp number to Claude as tools.

Here is what the WhatsApp group digest gives you:

• Decisions: what a group agreed while you were not reading.
• Asks you something: messages that name you or put a question to you.
• Dates and deadlines: anything with a day attached, with the group name.
• Not fully read: groups Claude could not read completely.

This is not the first WhatsApp MCP server. There are several, and we compare them in the best WhatsApp MCP servers.

How do you connect Claude to WhatsApp in under five minutes?

Link your WhatsApp number in Blueticks, then add https://api.blueticks.co/mcp to Claude as a custom connector and sign in to Blueticks to approve it. There is no API key to paste. The free Claude plan allows one custom connector, which is enough for this.

1. Link your number. Sign up at Blueticks and link WhatsApp the way you link WhatsApp Web, by scanning a QR code.
2. Open connectors in Claude. On Pro or Max the path is Customize > Connectors > + Add > Add custom connector (Anthropic's custom connector guide).
3. Add the server. Name it Blueticks and paste https://api.blueticks.co/mcp. Click Continue through the authentication settings, then click Add.
4. Approve. Claude sends you to a Blueticks sign-in screen. Sign in and approve. This is standard OAuth, so no key ends up in Claude.

On Team or Enterprise, an owner adds it under Organization settings > Connectors > Add > Custom > Web, per the same Anthropic guide. Custom connectors work on Claude Free, Pro, Max, Team, and Enterprise. Free is limited to one.

Using Claude Code or Cursor instead? The local npx @blueticks/mcp server with a bt_live_ API key still works. The step-by-step lives in how to connect WhatsApp to Claude, so I will not repeat it here.

The group digest prompt: what to paste and what Claude does

Paste one prompt into a normal Claude chat. Claude makes about three tool calls: it gets today's date, lists up to 10 group chats with unread messages from the last 3 days, then reads those groups in one batch. You get a digest grouped by decisions, asks, and dates, with the group name on every line.

Paste this as written:

Using Blueticks, make a digest of my WhatsApp groups. Get today's date first. Then list only group chats active in the last 3 days that have unread messages, including archived ones, max 10. Read their messages since then in one batch. Give me: decisions made, anything that names me or asks me something, and any dates or deadlines, with the group name on each line. List any group you could not read fully. Do not send, react, or mark anything as read.

What happens behind it, in plain words:

1. Date lookup. The server tells Claude to fetch the current date before any time-filtered call, so "last 3 days" means the right 3 days.
2. List groups. The chats tool can list group chats only. Each row carries an unread count, an archived flag, and the last message time. Archived chats are left out by default, which is why the prompt asks to include them.
3. Read in one batch. The list_messages action can read up to 25 chats in one call, 20 messages per chat by default when batched, with a since filter.

Approve each tool call when Claude asks. Do not choose "Allow always."

A sample digest, with invented groups and content:

Decisions

• Site team: delivery moved to the Thursday slot.
• Book club: next pick confirmed, meeting at Sam's.

Asks you something

• Product launch crew: "Can you send the final deck tonight?"
• Parents, Year 4: asks you to confirm the carpool rota.

Dates and deadlines

• Site team: invoice due on the 30th.
• Building committee: vote closes Monday at noon.

Not fully read

• Neighbours chat: truncated, too many messages in the window.

Check any date against the group before you rely on it. Claude reads message text and flags what names you or asks you something. It is a summary, not a record.

Does the digest fit in the free allowance?

Yes. On the free plan, the connector allows 5 requests per 6-hour window. Every tool call counts, including the date lookup. The digest is designed for about 3, and a follow-up uses a 4th. Windows reset at 00:00, 06:00, 12:00, and 18:00 UTC.

The count is shared across everything on your account: the connector and the REST API, and all your API keys. If Claude makes a stray extra call, you can hit the cap mid-task. Claude then gets a rate-limit error that says how long until the window resets.

Plans that include API access have no request quota, though per-key burst limits apply. One thing to know: the pricing page does not list this 5-request number. It does list a free plan with no time limit and no credit card, and says MCP is "included in every paid tier." The allowance above comes from the server's rate-limit policy.

Want the digest every morning without counting calls? Start free at blueticks.co/signup, send from your own number, and skip the API setup. Upgrade only if you outgrow 5 requests.

What breaks

The digest has four failure modes. Know them before you trust a quiet result.

• Engine offline. On free, Basic, and Standard, Blueticks runs in your browser, so WhatsApp Web must be open. If it is not, reads fail with 409 ENGINE_NOT_ONLINE. On Pro the engine runs in the cloud and keeps working when your computer is off (pricing).
• Partial coverage. A busy or cold group can come back truncated. The batch read has a 120-second time budget by default, and chats it could not reach are flagged as skipped. That is why the prompt asks Claude to list what it did not fully read.
• No muted filter. There is no "muted" field to filter on. The prompt picks groups by unread messages and recent activity. If you care about specific groups, name them.
• Similar names. "Site team" and "Site team 2" look alike to Claude. Use the exact group name.

How do you act on one item from the digest?

Pick one item and ask Claude to draft a reply or schedule a follow-up for it. Claude writes the draft, you read it, and you approve that single send. On the free plan this is your 4th request in the window. Everything before it was read-only.

1. Choose one line, for example the deck request from "Product launch crew."
2. Ask: "Draft a short reply to that message in the Product launch crew group. Show it to me first."
3. Edit the text in the chat until it says what you mean.
4. Approve the one send call. Decline anything else.

For catching up on one-to-one chats with the same draft-then-approve habit, see Claude as a WhatsApp personal assistant.

Does the connector ever send without you?

No. The connector does nothing on its own; it acts only when Claude calls a tool. The digest uses read actions only. But the same chats tool also holds send, react, mark-read, and archive actions, so approve each call yourself and run the digest in a normal chat, not Research mode.

Anthropic's own advice is to click "Allow always" only for a server and tool you trust to run unsupervised. It also says that in Research, Claude can invoke connector tools without asking (Anthropic's connector guide).

So three habits:

• Never pick "Allow always" for Blueticks.
• Run digests in a normal chat, not Research.
• Read the tool name and action before you approve. A list or list_messages call reads. A send_message_text call sends.

The prompt line "Do not send, react, or mark anything as read" is a request to Claude, not a lock. Your approval is the lock.

Need a summary of one long chat before a meeting instead?

Same connector, different prompt. Ask Claude to read one named chat and list open questions, dates, and who owes the next reply. That recipe has its own walkthrough in how to connect WhatsApp to Claude, which leads with it.

For example: "Summarize my chat with [exact name] from the last two weeks. Open questions, dates, who replies next. Do not send anything."

Which WhatsApp MCP tools does Claude get?

Standard accounts get nine tools, matching the developer docs page. Accounts approved for Claude agents also get agents and agent_schedules. The digest uses chats plus utils for the date. The chats tool reads and sends, which is why you approve each call.

How far back can it read? With a since date, the backend pages back through older history before it returns, and reports how far it got. A load_more_history action pulls older pages from the phone, up to 10 pages per call. No fixed maximum depth is documented.

WhatsApp MCP vs the WhatsApp Business API: which do you need?

Use the Blueticks WhatsApp MCP when you want Claude to read and act on your own WhatsApp number. It works through WhatsApp Web and the Blueticks engine, not the Meta Cloud API. If you need Meta's own business platform, that is a different product with its own onboarding.

Neither path comes with a guarantee against bans. If you send, consent and opt-out are your responsibility. The WhatsApp Business Messaging Policy allows contact only with people who gave you their number and agreed to hear from you. Hold yourself to that even on a personal account.

What are the limits and good practices?

Read freely, send deliberately. The free plan gives 5 connector requests per 6-hour window. Plans with API access have no request quota but do have per-key burst limits. Every send goes out from your real number, so WhatsApp's rules apply to each one.

• Read first, send one at a time. Keep the digest read-only. Approve sends individually.
• Name your groups. If a group matters, name it in the prompt rather than relying on unread counts.
• Keep the engine on. On browser plans, leave WhatsApp Web open while you run the digest.
• Check the plan. Prices per user per month on the pricing page:

Compare the plans on blueticks.co/pricing before you commit.

Frequently asked questions

Can Claude read WhatsApp groups? Yes. With the Blueticks connector, Claude can list your group chats and read their messages, up to 25 chats in one batch.

How far back can it read? It reads from the date you give it, paging back through older history, and can load more from the phone, 10 pages per call. No fixed maximum depth is documented.

Is it free? The digest fits the free plan's 5 requests per 6-hour window. Plans with API access have no request quota.

Does it send anything? Not on its own. It acts only when Claude calls a tool, and the digest uses read actions. Approve every call and skip "Allow always."

Does it work with my own number? Yes. You link your own WhatsApp number through the Blueticks engine, the same way you link WhatsApp Web.