Nimbata MCP connects your call data to Claude and ChatGPT, so you can tag calls, build workflows, read AI notes and get call reports by asking in plain English. Below are six ways marketers can use the connectors, each with prompts you can copy, plus what they can’t do yet.

Phone calls are often the most valuable leads a business gets. A caller has already decided they want to talk, and for many service businesses, clinics, agencies and B2B teams, a call converts far better than a form fill. That’s why marketers track them.

The tracking part is solved. The hard part is what comes after: turning thousands of call records into decisions.

Think about a normal week. On Monday, your manager asks which campaign brought the best calls last month. You open the call report, filter by date, switch the grouping to campaign, export a CSV, open it in a spreadsheet and build a pivot table to compare it with the month before. On Wednesday, sales says half of last week’s Google Ads calls were not real leads. To check, you open each call, listen to a few recordings and add tags by hand. On Friday, you want to know what callers keep asking about, so you can update a landing page. The answer is in the transcripts, but nobody has time to read 200 of them.

None of this is hard. It’s just slow, repetitive and easy to put off. We hit this ourselves recently, when we needed to tag 20 calls and the only way to do it was opening each one.

Nimbata MCP changes how this works. With the Nimbata connectors for Claude and ChatGPT, you ask a question or give an instruction in plain English, and the assistant reads your Nimbata data, does the work and hands you the answer:

“Which campaigns brought the most qualified calls last month, and how does that compare to September?”

Instead of five screens and a spreadsheet, you get a short table and a plain-language summary of what changed. The same goes for tagging calls in bulk, reading AI notes across dozens of calls, mining transcripts for insight, and looking back at months of history.

What is Nimbata MCP?

Nimbata MCP lets Claude or ChatGPT read and act on your Nimbata account when you ask in plain English. It’s built on MCP (Model Context Protocol), the open standard that lets AI assistants like Claude and ChatGPT work safely with the tools you already use. The examples in this post use Claude, but every prompt works the same way in ChatGPT.

In practice, it works like a very fast assistant who knows your Nimbata account inside out. You don’t need to know where a setting lives, which report has the answer or how a workflow is built. You describe the outcome you want, and Claude works out the steps.

Here’s what happens behind a single request like “Tag all paid search calls over 2 minutes from last week as Qualified Lead”:

  1. Claude finds the right project and the Qualified Lead tag.
  2. It searches last week’s calls and keeps the ones from paid search that lasted over 2 minutes.
  3. It shows you the matching calls so you can check them.
  4. Once you confirm, it tags each one and tells you when it’s done.

You wrote one sentence. Claude did the searching, filtering and clicking.

What the connectors can access

Area Can read Can change
Calls Call details, recordings, transcripts, AI notes and call reports A call’s tags, notes, rating, value and custom fields
Tags and fields All tags and custom fields Create and edit tags and custom fields
Workflows and call flows Every workflow and call flow, and what it does Create and edit them (new workflows start switched off)
Alerts Email alerts, Slack alerts and webhooks Create, edit and delete them
Numbers and sources Tracking numbers, tracking sources and blocked callers Tracking sources and blocked callers (add only)

It can also copy a setup, like tags, sources and blocked callers, from one project to another. What it can change depends on the access level you choose when you connect (see View or View and manage).

1. Qualify, label and value calls without the clicking

A raw call record tells you who called, when, and from which campaign. To make it useful for marketing, you add the rest: was it a real lead, what happened on the call, and what it was worth. In Nimbata, you do that with tags, call values, custom fields and workflows.

That extra context matters more than it seems. It decides which calls count as conversions in Google Ads, which campaigns look good in your reports, and whether sales trusts the leads marketing sends them.

The tools are all there. The effort is in using them: finding the right calls, deciding what each one was, building the rules and keeping them tidy. That’s the part Claude takes off your plate.

Update a batch of calls in one sentence

Today: you filter the call list by date and source, scan the results for the calls that fit, then open each one, add the tag and save. If the filter you need doesn’t exist, like call duration, you check each call by eye. Remember our 20 calls? That was 20 rounds of open, tag, save, and it took [X] minutes by hand. With MCP, it took one prompt and a one-minute check.

With MCP:

  • “Tag all paid search calls over 2 minutes from last week as Qualified Lead.”
  • “Tag every unanswered Google Ads call from yesterday as Missed Opportunity.”
  • “Tag all first-time callers from the Spring Promo campaign as New Customer.”
  • “Remove the Spam tag from all calls from this number.”

What it looks like: Claude replies with the calls it found, for example “I found 23 paid search calls over 2 minutes between November 2 and 8”, along with the date, caller and duration of each. You confirm, and it tags them all. If a few don’t belong, tell it which ones to skip.

Why it’s easier: you describe the calls instead of hunting for them, and a ten-minute chore becomes a one-minute check.

Pro tip: tagging can trigger your workflows. If a tag like Qualified Lead sends an offline conversion to Google Ads, a bulk tag sends all of them. That’s a fast way to fix missed conversions, so just make sure it’s what you want. Keep in mind that Google Ads only accepts offline click conversions within 90 days of the ad click, so calls from older clicks won’t import.

Let Claude do the listening and the judging

Today: deciding if a call was a real lead, a booking or a sales pitch means playing the recording or reading the transcript yourself. Duration helps a little, but a 3-minute call can be a supplier pitch, and a 50-second call can be a booked job.

With MCP:

  • “Read yesterday’s transcripts and tag each call Qualified, Not a Fit or Spam. Show me your reasoning before you tag.”
  • “Which of last week’s untagged calls sound like booked jobs?”
  • “Find the calls where the caller mentioned a competitor and tag them Competitor.”
  • “Which calls from this week were existing customers asking for support, not new leads?”

What it looks like: a short list of calls, each with Claude’s suggested label and a one-line reason, such as “Caller asked for a quote for a bathroom remodel and gave an address”. You approve the list, edit the odd one, and Claude applies the tags.

Why it’s easier: Claude does the first pass. You review its decisions instead of doing the work from scratch, and your reports finally separate real leads from noise.

Build workflows by describing them

Today: setting up a workflow means picking a trigger, adding conditions, choosing actions and mapping fields, step by step. You need to know where each option lives and how they fit together, so workflows often stay on the to-do list.

With MCP:

  • “When a Google Ads call lasts over 90 seconds, tag it Qualified Lead and send it to Google Ads as a conversion.”
  • “Create a workflow that sets the call value to 150 when a call is tagged Booked.”
  • “Add a Service field with Repair, Install and Maintenance, and have AI fill it after every call.”
  • “Fill the call value from the matching HubSpot deal.”

What it looks like: Claude builds the workflow switched off, then explains it back in plain words: what triggers it, what it checks and what it changes. You read the summary, ask for tweaks if needed, and publish when it looks right.

Why it’s easier: you say what you want in one sentence instead of learning the builder. This is also the permanent fix for bulk tagging: once the rule is live, new calls get labeled automatically.

Check and tidy your setup in one conversation

Today: knowing which tags are in use, which workflows are running and why a call wasn’t tagged means opening one screen after another, and comparing notes across projects.

With MCP:

  • “Which tags did we use last month, and which ones are never used?”
  • “List our active workflows and explain in plain words what each one does.”
  • “Why didn’t this call get tagged Qualified Lead?”
  • “Copy our tags, fields and workflows to the new client’s project.”

What it looks like: a plain-language audit, for example a list of workflows with one line each on what they do, plus the tags nobody has used in months.

Why it’s easier: a full health check takes one conversation, and every project stays set up the same way. For agencies, that consistency makes cross-client reporting much simpler.

2. Get answers from AI notes without reading them one by one

Nimbata already writes AI notes on your calls: summaries, action plans, coaching notes and smart follow-ups. They’re a huge time saver on a single call. You no longer need to listen to a ten-minute recording to know what happened.

The catch is that each note sits on its own call. When you want the bigger picture, like what happened across all of yesterday’s calls or which leads still need a follow-up, you’re back to opening calls one at a time. Claude reads across all of them at once and turns them into answers.

See the big picture across many calls

Today: to spot a pattern, you open call after call and read each note. With 30 calls a day, that’s a morning gone, so most teams skip it.

With MCP:

  • “Summarize the AI notes from yesterday’s calls in five bullets for the morning standup.”
  • “Which follow-ups from this week’s calls are still open?”
  • “List every call this week where the action plan says to send a quote.”
  • “What were the most common reasons people called this week?”

What it looks like: a short brief you can paste into Slack or read out in a meeting, for example how many calls came in, the main topics, the hottest leads and the follow-ups due today.

Why it’s easier: a morning of reading becomes one question with a short answer, and nothing slips through because nobody had time to check.

Get the story of one caller in seconds

Today: before you call someone back, you search for their number, find every call they made and read through each one. Repeat callers are the hardest, because the context is spread across several calls.

With MCP:

  • “What did the caller from +30 210 000 0000 want, and what did we promise them?”
  • “Summarize every call from this number in the last 30 days.”
  • “Has this caller called before, and from which campaign did they first find us?”

What it looks like: a few lines covering who the caller is, what they asked for, what was agreed and which campaign or page first brought them in.

Why it’s easier: you go into the callback with full context, without digging. For marketers, it also shows the real path a lead took before converting.

Set up and fine-tune AI notes in a sentence

Today: turning AI notes on for a project, or changing what they capture, means building or editing workflows by hand and writing the AI instructions yourself.

With MCP:

  • “Turn on AI summaries and Smart Follow-up notes for all answered calls in this project.”
  • “Change the AI note so it also records the budget the caller mentioned.”
  • “Send each call’s AI summary to our Slack sales channel.”
  • “Add the same AI note setup to all our client projects.”

What it looks like: Claude sets up the workflow, explains what each note will capture, and leaves it switched off until you publish it.

Why it’s easier: you describe the notes you want, and Claude handles the setup. You can shape the notes around what your team actually needs, like budget, timeline or the service requested.

3. Turn call transcripts into marketing insight

Your callers tell you exactly what they want, what worries them and how they describe their problem, in their own words. That’s the best voice-of-customer research you’ll find. Surveys and keyword tools only guess at it.

Most of it sits unread in transcripts, because reading 200 calls is nobody’s job. Claude can read weeks of them in one go and pull out what matters for marketing. Here’s what to look for and where each insight pays off.

Insight Prompt Use it in
Top questions “Give me the top 5 questions from last week’s calls.” FAQ blocks and landing pages
Objections “Which objections come up most on calls from the Brand campaign?” Ad copy and pricing pages
Caller wording “What words do callers use to describe their problem?” Keywords and headlines
Competitor mentions “Which competitors do callers mention most, and what do they say?” Messaging and budget
Service demand “Which services did people ask about most this month?” Budget and campaign priorities
Campaign fit “Are callers from this campaign asking for what the ads promise?” Negative keywords and ad targeting

Here are five ways to put it to work.

Find the questions to answer on your pages

  • “Read last week’s answered call transcripts and give me the top 5 questions people ask.”
  • “Which questions do callers ask that our pricing page doesn’t answer?”

How to use it: add the answers to your landing pages and FAQ sections. Every question a page answers is one less reason to bounce, and good material for content that AI search engines can quote.

Example: on a plumbing company’s project, asking for the top 5 questions from last month’s answered calls could return:

  1. “Do you charge a call-out fee?”
  2. “Can someone come out today?”
  3. “Will you give me a price before you start?”
  4. “Are you licensed and insured?”
  5. “Do you work weekends?”

If the landing page only answers the last two, the fix is an FAQ block on call-out fees, same-day visits and upfront quotes, written in the callers’ own words.

Learn the objections and write around them

  • “Which objections come up most on calls from the Brand campaign?”
  • “Why do callers say no after hearing the price?”

How to use it: address the top objections in your ad copy and on the page before the call happens, with guarantees, financing, reviews or clearer pricing.

Borrow your customers’ words

  • “What words do callers use to describe their problem? List them so I can compare with our keywords.”
  • “How do callers describe the service they want, in their own words?”

How to use it: compare the list with your keyword lists and ad headlines. Callers often use different words than marketers do, and those are new keyword and headline ideas.

Track competitors and demand

  • “Find the calls that mention pricing or a competitor, and tag them.”
  • “Which competitors do callers mention most, and what do they say about them?”
  • “Which services did people ask about most this month?”

How to use it: spot which competitors you’re compared with, and which services are in demand, so you can shift budget and messaging to match.

Check that campaigns attract the right people

  • “Are callers from the Emergency Repair campaign actually asking for emergency repairs?”

How to use it: if callers from a campaign ask for something else, the keywords or ad copy are pulling the wrong audience. That’s a quick win for negative keywords.

Why marketers care: call transcripts turn into ad copy, landing page content, FAQs, keyword ideas and competitor intel, all in the words real customers use.

4. Get reports without exports or pivot tables

Nimbata’s reports already hold the numbers. What takes time is getting a specific answer out of them: picking the right report, setting the filters, exporting and comparing periods in a spreadsheet.

With Claude, you skip straight to the question. You get the breakdown, a comparison to the previous period when you need one, and a plain-language summary of what changed. The best part is the follow-up. When an answer raises a new question, like “why did that campaign drop?”, you just ask it, and Claude digs one level deeper without you starting over.

Campaign and channel performance

  • “Which campaigns drove the most calls last month, compared to the month before?”
  • “Break down calls by source / medium for Q3.”
  • “Compare calls from Google Ads, Bing and Facebook this month.”

The decision it supports: where to put next month’s budget, and which campaigns or channels to scale or pause.

Lead quality and value

  • “How many Qualified Lead calls did each campaign bring in October?”
  • “Which campaigns bring the highest total and average call value?”
  • “How many new vs repeat callers came from Paid Search this quarter?”

The decision it supports: which campaigns bring real customers, not just call volume. A campaign with fewer calls but more qualified leads or higher value is often the one to grow.

Landing pages and caller journeys

  • “Which landing pages generated the most calls last month?”
  • “Which pages did callers browse before calling from Google Ads?”
  • “Do people call more from the pricing page or the contact page?”

The decision it supports: which pages to test, improve or send more traffic to, based on the pages that actually make people pick up the phone.

Call handling

  • “What share of calls went unanswered this week, by tracking number?”
  • “What’s the average call duration by tracking source?”
  • “What time of day and which weekday do Paid Search calls peak?”

The decision it supports: ad scheduling and staffing. If calls peak at times nobody answers, you’re paying for clicks that end in voicemail.

  • “Show daily calls for the last 30 days and flag any unusual drops.”
  • “Did calls from any campaign drop more than 20% vs last month?”
  • “Which tracking numbers got no calls this week?”

The decision it supports: catching problems early. A sudden drop often means a broken number, a missing tracking script or a paused campaign.

Why marketers care: budget and bid decisions get made on calls and call quality in minutes, not hours. Client and management reports take one question instead of an afternoon. And broken tracking or missed calls get caught before they cost you a week of ad spend.

5. Look back at months of calls

Some of the best questions only come up after the fact. Did the new landing page work? When did lead quality start slipping? What were people asking for last summer?

Answering them usually means rebuilding old reports or giving up because the data was never tagged properly. Claude can go back through months of call history and answer questions you didn’t think to ask at the time.

Measure the impact of a change

  • “Compare calls in the 30 days before and after we launched the new landing page on October 1.”
  • “Did qualified calls from Google Ads change after we switched to a new bidding strategy in September?”

How to use it: prove (or disprove) that a change worked, with call data instead of gut feeling.

Ask new questions of old calls

  • “Go through Q3 transcripts and tell me which services people asked about most, by month.”
  • “When did callers start mentioning our new competitor, and from which campaigns?”

How to use it: spot trends that started months ago, and plan campaigns and content around them.

Fix gaps in past data

  • “Re-qualify last month’s untagged Paid Search calls using our new Qualified Lead rules.”
  • “Tag last quarter’s calls that mentioned a booking as Booked.”

How to use it: when you add a new tag or rule, apply it to past calls too, so your reports compare like with like. If those tags feed Google Ads, this can also recover conversions that were never sent, but only for calls whose ad click was in the last 90 days. Google Ads doesn’t accept offline conversions for older clicks, so tagging calls from earlier months fixes your Nimbata reports without changing Google Ads.

Compare seasons and years

  • “How did call volume and value by campaign change this year vs last year?”
  • “Which campaigns brought the callers who called back more than once in the last 90 days?”

How to use it: plan budgets around seasonality, and find which campaigns bring loyal customers, not just first calls.

Why marketers care: you can measure the real impact of a change, fix gaps in past data, and make a stronger case for budget with history to back it up.

6. Alerts, agencies and account setup

The everyday admin work around call tracking gets shorter too. These are small tasks on their own, but they add up across a week, and even more across a portfolio of clients.

Get the right people alerted in real time

  • “Email me and the sales manager whenever a Google Ads call is missed.”
  • “Post a Slack alert in #sales when a call is tagged Hot Lead.”
  • “Email the account manager when a call from our top campaign lasts over 5 minutes.”

Why it helps: a missed call from paid traffic is money already spent. Fast alerts mean sales calls back while the lead is still warm.

For agencies: run every client project the same way

If you’re an agency, each client usually has its own Nimbata project. The assistant sees every client project your role can reach, so you can set up, compare and report on all of them from one chat instead of opening each account.

  • “Copy the tags, workflows and spam list from Client A to Client B.”
  • “Set up the new client’s project with the same tracking sources as our template project.”
  • “Compare call volume and qualified calls across all my client projects for October.”
  • “Which client had the biggest drop in calls this month, and from which campaign?”

Why it helps: onboarding a new client takes minutes instead of an afternoon, every client account follows your agency’s standard setup, and monthly client reports and account handovers get much easier. See how Nimbata works for agencies.

Keep tracking and routing healthy

  • “Is our tracking code installed and swapping numbers on the site?”
  • “List all tracking numbers and which source each one belongs to.”
  • “Build a call flow: business hours go to sales, after hours go to voicemail.”
  • “Block this caller in all projects.”

Why it helps: you check that tracking works after every site change, route calls without learning the call flow builder, and keep spam out of your data across every project.

Why marketers care: sales hears about hot leads in real time, onboarding a client takes minutes, and setup checks no longer need five screens.

Prompt cheat sheet

Copy these into Claude or ChatGPT and swap in your own campaigns, tags and dates.

Job Prompt to copy
Bulk-tag calls Tag all paid search calls over 2 minutes from last week as Qualified Lead
AI qualification Read yesterday’s transcripts and tag each call Qualified, Not a Fit or Spam
Build a workflow When a Google Ads call lasts over 90 seconds, tag it Qualified Lead and send it to Google Ads
Setup check List our active workflows and explain what each one does
AI notes digest Summarize the AI notes from yesterday’s calls in five bullets
Callback prep What did the caller from this number want, and what did we promise them?
Voice of customer Top 5 questions from last week’s call transcripts
Campaign performance Which campaigns drove the most calls last month vs the month before?
Lead quality How many Qualified Lead calls did each campaign bring in October?
Call value Which campaigns bring the highest total and average call value?
Landing pages Which landing pages generated the most calls last month?
Missed calls What share of calls went unanswered this week, by tracking number?
Trend check Show daily calls for the last 30 days and flag any drops
Before and after Compare calls 30 days before and after our new landing page went live
Missed-call alert Email me when a Google Ads call goes unanswered
Client onboarding Copy tags, workflows and spam list from Client A to Client B
Tracking check Is our tracking code installed and swapping numbers on the site?

View or View and manage

When you connect Nimbata to Claude or ChatGPT, you pick one of two access levels.

Permission What Claude or ChatGPT can do Best for
View Look up calls, transcripts, AI notes, reports and settings. It can’t change anything in Nimbata. Reporting, analysis and anyone who only needs answers. A safe way to try it first.
View and manage Everything in View, plus tag calls, set values and fields, and create or edit tags, workflows, call flows, alerts and webhooks. Marketers and admins who run the account, bulk tagging and agency setups.

At either level, buying or releasing numbers, billing, users and roles, sending text messages and removing a blocked caller stay in the Nimbata app. Admins choose which roles can connect, and at which level, with the MCP Access permission.

What the connectors can’t do (yet)

A few things still happen in the Nimbata app, and a few depend on the data you collect.

  • Buy or release phone numbers. Do this in the Nimbata app.
  • Change billing. Also in the app.
  • Send text messages. Claude can read your SMS logs, but it can’t send messages.
  • Filter every field directly. Some filters, like call duration, aren’t built into call search. Claude pulls the calls and filters them itself, which works well for weekly batches. For ongoing rules, a workflow is the better tool.
  • Create data you don’t collect. Without transcription there are no transcript insights, and calls without UTMs show up as “(not set)”.

Tips for getting the best results

  1. Preview before bulk changes. Ask “show me the list first, then tag them” so you can check the matches.
  2. Know that changes go live right away. Tagging a call can trigger your workflows, for example sending an offline conversion to Google Ads or a lead to your CRM. That’s often exactly what you want, but it’s good to know.
  3. Be specific. Name the project, the date range and the exact tag.
  4. Reports cover inbound calls. Outbound calls show up in call search, not in call reports.
  5. Review new workflows before they run. Claude creates them switched off, so you can check them and publish when ready.

FAQ

What is Nimbata MCP?

It’s a secure connection between your Nimbata account and AI assistants like Claude and ChatGPT, built on the Model Context Protocol (MCP). Once connected, you can ask questions about your calls and make changes in Nimbata from a chat.

Does Nimbata MCP work with ChatGPT?

Yes. Nimbata connects to ChatGPT the same way it connects to Claude, and every prompt in this post works in both. ChatGPT can look up your calls, call reports, sources, tags, workflows and settings, and with View and manage access it can make changes too.

How do I connect Nimbata to Claude or ChatGPT?

Add a connector in Claude or ChatGPT with the server URL https://mcp.nimbata.com/mcp, sign in with your Nimbata login and choose View or View and manage. There’s no code to write and no API key to copy. In ChatGPT, turn Nimbata on in a new chat. On Claude Team and Enterprise, an Owner adds Nimbata for the organization first.

What's the difference between View and View and manage?

With View, the assistant can look things up but can’t change anything in Nimbata. With View and manage, it can also tag calls, set values and fields, and create or edit tags, workflows, call flows, alerts and webhooks.

Can Claude tag calls in Nimbata?

Yes. Describe the calls you want, like “all paid search calls over 2 minutes from last week”, and the tag to add. Claude finds the matching calls, shows you the list and tags them.

Can I bulk-tag calls in Nimbata?

Yes, through Claude. One instruction can tag dozens of calls at once, instead of opening each call by hand.

Can Claude analyze my call recordings and transcripts?

Yes, for calls that have transcription. Claude can read transcripts to qualify calls, find common questions and objections, or summarize what callers wanted.

Can Claude buy tracking numbers or send SMS?

No. Buying or releasing numbers, billing and sending text messages are done in the Nimbata app. Claude can read your SMS logs.

Is it safe to let AI change my call tracking settings?

Claude only changes what you ask for. New workflows are created switched off so you can review them, and you can always ask Claude to show you what it will change before it acts.

Do I need to be technical to use Nimbata MCP?

No. If you can describe what you want in a sentence, you can use it.

Does the assistant see everything in my Nimbata account?

No. It works as you, so it sees only the projects and data your Nimbata role can reach. Admins decide which roles can connect, read-only or with full access, through the MCP Access permission.

Can agencies use it across client projects?

Yes. The assistant can read every project your role can reach, so you can compare clients in one question and, with View and manage, copy tags, sources and blocked callers from one project to another.

Which Nimbata plans include MCP?

Every Nimbata plan, at no extra cost. You also need a Claude or ChatGPT account that can add connectors.

How do I disconnect Claude or ChatGPT?

In Nimbata, open My Profile, then Connected apps, and click Disconnect. It takes effect right away. A connection also ends after a year, after 90 days without use, or when you change your password.

Try it on your own calls

Connect Nimbata to Claude or ChatGPT in a few clicks and start with one prompt from the cheat sheet. Pick the task you dread most this week, whether that’s tagging, reporting or reading call notes, and see how long it takes with one sentence.