InterLogue

Brief it like a reporter.
It phones them and writes the story.

An MCP tool inside Claude that interviews your customer by phone and returns a publishable case study with every quote cited to a timestamp in the transcript.

No account. No API key. Runs on a fictional fixture until you approve a real call.

Add it to Claude in six steps.

Copy the connector address, then follow the steps in claude.ai or Claude Desktop.

https://interlogue.onrender.com/mcp
  1. Open claude.ai or Claude Desktop, then Settings, then Connectors.
  2. Choose Add custom connector.
  3. Name it InterLogue and paste the address above.
  4. Leave authentication empty and click Add.
  5. Start a new chat with InterLogue switched on.
  6. Paste the sample brief below and say the person may be contacted.

You should see a piece with a timestamp on every quote.

The hosted server is a shared, text-only sandbox: use fictional names. Phone interviews need the Desktop install.

The sample brief

Interview Marisol Teague, +1-502-555-0142, founder of Ridgeline Provisions, about switching her order entry to Tallyhook, for Tallyhook's marketing team. I approve contacting her at that number. Use the founder-case-study fixture instead of calling, then write the piece and show it to me.

Or install it in Claude Desktop

  1. Download interlogue.mcpb (3.1 MB).
  2. Double-click the file; Claude Desktop opens the install dialog.
  3. Confirm. The InterLogue tools appear in the tools menu.
  4. Paste the same brief.

Claude Desktop supplies the Node runtime. Nothing in the bundle needs credentials.

One interview, measured.

6 min 12 s

From brief to published piece, in the rehearsal

80¢

Platform cost per ten minutes of call

100%

Of quotes cited to a transcript timestamp, or the piece does not publish

58

Tests, including tamper tests on quotes and timestamps

Built with

Claude Model Context Protocol ElevenLabs Twilio Node.js

The story comes first. Every quote carries its timestamp.

Fictional fixture. Marisol Teague, Ridgeline Provisions and Tallyhook are invented. This is the piece Claude wrote from the repo's text fixture, exactly as InterLogue published it: story first, pull quotes with timestamps second, the verbatim per-question record third.

“We got our Mondays back”: Ridgeline Provisions on the end of hand-typed orders

A customer case study with Marisol Teague, founder of Ridgeline Provisions, for Tallyhook.

Marisol Teague is the founder of Ridgeline Provisions. “Ridgeline Provisions is a specialty food wholesaler. We buy from about forty small producers, mostly within a couple of hours of us, and we sell to restaurants and independent grocers,” Teague said. (00:31) The operation is small, and Teague does not dress that up. “It's really just the two of us on the operations side. Theo runs the warehouse and the routes, and I do purchasing, sales, and until last year, all of the order entry.” (01:15) “So when people say small team, I mean small. If one of us is sick, the other one is doing both jobs.” (01:15)

Ask about the old process and what comes back is a description of Mondays. “So orders came in every way you can imagine. Chefs would text me at eleven at night. Some of the grocers emailed a spreadsheet. A couple of the older restaurants literally called and read their order to me over the phone. And every single one of those I typed by hand into a Google Sheet.” (01:52) Then the worst day of the week: “Monday was the worst, because everybody orders for the week on Sunday night. I'd sit down at six in the morning with maybe sixty orders across texts, emails and voicemails, and I'd be keying them in until lunch. Six hours, most Mondays, just typing.” (01:52)

The hours were only part of it. “The mistakes are what actually hurt. If I misread a text and sent a restaurant two wheels of cheese instead of two wedges, that's a hundred and forty dollars of product on a truck going the wrong way, and a chef who is annoyed with us,” Teague said. (02:46) The arithmetic had been done. “I sat down once and figured we were eating around two thousand dollars a month in wrong-order credits and wasted product. And I was working every Sunday night because I was scared of Monday.” (02:46)

The turning point was a specific week, and Teague tells it in order. “There was a specific week last spring. I had the flu, Theo did the Monday entry, and he made a bunch of errors because he didn't know the customers' shorthand. We lost an account over it, a restaurant we'd had for four years. That was the week I said okay, this cannot depend on what is in my head.” (03:45) The search that followed is told just as briskly. “We looked at three things. One of the big enterprise systems, which quoted us something like eighteen thousand a year and wanted a six-month implementation. A generic inventory app that didn't really understand wholesale ordering. And Tallyhook, which a produce guy I know had mentioned.” (03:45)

The reason Teague gives for the choice is not a feature list. “Honestly, the customer-facing side. Tallyhook gave every customer a little order page, and it also reads the emails and the texts and turns them into orders that I approve instead of type. That was the thing. I didn't need to change how my chefs behave, because chefs are not going to change. They're still texting me at eleven at night. The difference is now it lands as a draft order that I glance at and tap yes.” (04:44) On price: “And the price was three hundred a month, which against two thousand a month in mistakes was not a hard conversation to have with Theo.” (04:44)

Teague is candid about the first weeks. “Messier than the sales pitch, I'll be honest. The first week I ran both systems side by side because I didn't trust it. I'd let Tallyhook draft the orders and then I'd check every one against what I would have typed. It caught two things I would have gotten wrong, and I caught one it got wrong, a customer who says 'the usual' and expects us to know.” (05:36) “By the second week I stopped double-entering. By the third week Theo was pulling pick lists straight out of it and the printer in the warehouse had basically stopped running.” (05:36)

The result Teague leads with is a number. “Monday order entry went from about six hours to about forty minutes. That's the one I tell people. Forty minutes, and most of that is me reading, not typing,” Teague said. (06:37) The rest follow. “Wrong-order credits went from around two thousand a month to under three hundred, and the ones we still get are usually the producer's fault, not ours. We've added eleven new accounts since the spring, which is more than we added in the two years before that, because I actually have time to go sell. And I don't work Sunday nights anymore. That one doesn't have a number on it, but it's the one that matters to me.” (06:37)

There is also a moment, and it is the one in the headline. “The second Monday after we switched, I finished the orders, looked at the clock, and it was seven fifteen in the morning. I actually thought something was broken. I went out to the warehouse to ask Theo if the orders had come through, and he already had the first truck half loaded. I stood there for a second and thought, we got our Mondays back. That's still how I describe it to people.” (07:36)

Teague's advice to another founder starts with the cost of waiting. “Don't wait for the disaster. I waited for the disaster, and the disaster cost me a four-year account. If you're the only person who can do the order entry, you don't have a process, you have a hostage situation, and the hostage is you. Also, run it side by side for a week if you're nervous. It costs you one extra week and you'll actually believe the numbers after that.” (08:20)

Pull quotes

“Six hours, most Mondays, just typing.” (01:52)
“I sat down once and figured we were eating around two thousand dollars a month in wrong-order credits and wasted product.” (02:46)
“I didn't need to change how my chefs behave, because chefs are not going to change.” (04:44)
“Monday order entry went from about six hours to about forty minutes. That's the one I tell people.” (06:37)
“I stood there for a second and thought, we got our Mondays back.” (07:36)
“If you're the only person who can do the order entry, you don't have a process, you have a hostage situation, and the hostage is you.” (08:20)
Verbatim record: per-question answers

Thanks. This should take about ten minutes, and you can stop me at any time. To start, can you tell me a bit about Ridgeline Provisions and what you do there? (00:22)

Sure. Ridgeline Provisions is a specialty food wholesaler. We buy from about forty small producers, mostly within a couple of hours of us, and we sell to restaurants and independent grocers. Cheese, cured meats, preserves, some produce in season. I started it six years ago out of the back of a van, honestly. Today it's me and my business partner Theo, plus two drivers who are part-time.

Answer at 00:31

How big is the team day to day, and who handles the operations side? (01:08)

It's really just the two of us on the operations side. Theo runs the warehouse and the routes, and I do purchasing, sales, and until last year, all of the order entry. The drivers are on the road, they don't touch the office stuff. So when people say small team, I mean small. If one of us is sick, the other one is doing both jobs.

Answer at 01:15

Before Tallyhook, how did Ridgeline handle order entry and fulfillment? Walk me through what a typical week looked like. (01:44)

So orders came in every way you can imagine. Chefs would text me at eleven at night. Some of the grocers emailed a spreadsheet. A couple of the older restaurants literally called and read their order to me over the phone. And every single one of those I typed by hand into a Google Sheet. Monday was the worst, because everybody orders for the week on Sunday night. I'd sit down at six in the morning with maybe sixty orders across texts, emails and voicemails, and I'd be keying them in until lunch. Six hours, most Mondays, just typing. Then Theo would print the sheet and build the pick lists off it by hand.

Answer at 01:52

What did that cost you, in time, money or mistakes? (02:40)

The time part I've already said, call it six hours every Monday, and probably another eight across the rest of the week. The mistakes are what actually hurt. If I misread a text and sent a restaurant two wheels of cheese instead of two wedges, that's a hundred and forty dollars of product on a truck going the wrong way, and a chef who is annoyed with us. We were doing that maybe three or four times a week. I sat down once and figured we were eating around two thousand dollars a month in wrong-order credits and wasted product. And I was working every Sunday night because I was scared of Monday.

Answer at 02:46

What made you start looking for something different, and what else did you consider? (03:38)

There was a specific week last spring. I had the flu, Theo did the Monday entry, and he made a bunch of errors because he didn't know the customers' shorthand. We lost an account over it, a restaurant we'd had for four years. That was the week I said okay, this cannot depend on what is in my head. We looked at three things. One of the big enterprise systems, which quoted us something like eighteen thousand a year and wanted a six-month implementation. A generic inventory app that didn't really understand wholesale ordering. And Tallyhook, which a produce guy I know had mentioned.

Answer at 03:45

Why did you pick Tallyhook in the end? (04:38)

Honestly, the customer-facing side. Tallyhook gave every customer a little order page, and it also reads the emails and the texts and turns them into orders that I approve instead of type. That was the thing. I didn't need to change how my chefs behave, because chefs are not going to change. They're still texting me at eleven at night. The difference is now it lands as a draft order that I glance at and tap yes. And the price was three hundred a month, which against two thousand a month in mistakes was not a hard conversation to have with Theo.

Answer at 04:44

What did the first few weeks with Tallyhook actually look like? (05:30)

Messier than the sales pitch, I'll be honest. The first week I ran both systems side by side because I didn't trust it. I'd let Tallyhook draft the orders and then I'd check every one against what I would have typed. It caught two things I would have gotten wrong, and I caught one it got wrong, a customer who says 'the usual' and expects us to know. We set up a standing order for him and that was that. By the second week I stopped double-entering. By the third week Theo was pulling pick lists straight out of it and the printer in the warehouse had basically stopped running.

Answer at 05:36

What has changed since? If you have numbers, I would love to hear them. (06:30)

Monday order entry went from about six hours to about forty minutes. That's the one I tell people. Forty minutes, and most of that is me reading, not typing. Wrong-order credits went from around two thousand a month to under three hundred, and the ones we still get are usually the producer's fault, not ours. We've added eleven new accounts since the spring, which is more than we added in the two years before that, because I actually have time to go sell. And I don't work Sunday nights anymore. That one doesn't have a number on it, but it's the one that matters to me.

Answer at 06:37

Is there a moment where you noticed the difference? (07:30)

Yeah. The second Monday after we switched, I finished the orders, looked at the clock, and it was seven fifteen in the morning. I actually thought something was broken. I went out to the warehouse to ask Theo if the orders had come through, and he already had the first truck half loaded. I stood there for a second and thought, we got our Mondays back. That's still how I describe it to people.

Answer at 07:36

What would you tell another founder in your position who is weighing this? (08:14)

Don't wait for the disaster. I waited for the disaster, and the disaster cost me a four-year account. If you're the only person who can do the order entry, you don't have a process, you have a hostage situation, and the hostage is you. Also, run it side by side for a week if you're nervous. It costs you one extra week and you'll actually believe the numbers after that.

Answer at 08:20

Is there anything you would want the Tallyhook team to hear, or anything I should have asked? (09:00)

Only that their support people actually answer the phone. And no, I think you got it.

Answer at 09:08

That is everything I have. Thank you, Marisol. The Tallyhook team will send you the draft before anything is published, and nothing goes out without your okay. (09:26)

Great, thanks. Bye now.

Answer at 09:40

Consent

Agent (00:00): Hi, is this Marisol? This is the InterLogue interviewer calling on behalf of the marketing team at Tallyhook. Before we get going, two things. I am an AI, not a person, and I would like to record this call so that anything we quote is accurate to what you actually said. Is it okay with you if we record?

Marisol Teague (00:14): Yes, that's fine. Devon at Tallyhook said you would be calling. Go ahead.

Citation check: OK, 23 quotes resolved to transcript timestamps.

What it tells you.

InterLogue advises; it does not get in the way. Two notices you will see, as the tools print them today.

When the interview was thin

The interview is graded before the piece is written. A short or evasive call still gets its packet; the notice sits at the top and the headline must say the interview was brief. It is advice, not a block.

## Thin interview notice (advice, not a block)
This interview graded below the story threshold. Write the piece anyway, keep it as short as the material, and say in the headline that the interview was brief. If the user wants more, the fix is a second interview asking the unanswered beats first.
- substantive subject words 80, minimum 120
- planned beats with a real answer 2 of 7, minimum 4
- quote candidates of 8+ words 4, minimum 6
Beats without a real answer: context, search, implementation, results, reflection.

When the subject is the founder

A customer case study needs a customer at another company. When the subject's company is the client, the brief switches to a founder story and asks about what they built instead of "before the product". Nothing is refused; the plan changes.

BRIEF SAVED: brf_mtyo4ahx6acb2d
Genre: founder story (the subject's company is the client, so the plan asks about what they built)
Subject: Jane Doe, founder, a small bakery in Louisville (phone ending 0100)

Three rules.

  1. A human approves every person and number before a dial.
  2. The agent says it is an AI and asks to record before anything else.
  3. Every quote resolves to a transcript timestamp, or the piece does not publish.

Questions.

Does it call anyone I name?

No. A human records approval of the specific person and number first, in their own words, and nothing dials without that record. The hosted connector is text-only; real calls need the Desktop install and your own credentials.

Does the subject know it is an AI?

Yes. The agent says it is an AI and asks permission to record before it asks anything else. Any answer short of an explicit refusal counts as a yes, and an explicit refusal stops the piece.

Can a quote be invented?

No. Before a piece is stored, every quoted span is checked verbatim against the subject's words at the cited timestamp. A quote that does not resolve fails the check, and nothing is published until it does.

What does an interview cost?

About seven cents per minute of call as reported by the voice platform on our account, so under a dollar for ten minutes, against two hours of a marketer's time. The carrier's per-minute charge is separate.

What happens if the interview is bad?

The interview is graded before the piece is written. A thin one still gets written, as short as the material, with a headline that says the interview was brief, and the packet names the beats that went unanswered so a second interview can ask them first. Nothing goes public without a human review.