Lindy bills 20 credits per minute and writes orders to HubSpot or Slack. A restaurant phone order needs to land on the POS.
Lindy AI Phone Agent (Gaia) is a horizontal voice runtime that bolts onto Lindy's 5,000+ business-tool directory. The directory is full of CRMs, productivity tools, and Slack. It is not full of restaurant POS adapters. This guide walks the per-call cost on a real ordering line, the integration shape Gaia actually has, and what happens to a half-pepperoni-half-mushroom pizza when a horizontal phone agent answers the published number.
What Lindy actually is, in plain terms
Lindy is an agent platform from Lindy.ai. You build agents in a no-code canvas, give them a knowledge base, and connect them to whatever business tools you already use. Gaia is the phone product layered on top: it lets a Lindy agent answer inbound calls and place outbound calls, with the same knowledge and the same integrations as the rest of the agent. The voice runtime is horizontal, meaning it is not built for any one industry. Sales calls, support calls, lead qualification, scheduling, surveys, and back-office workflows all run on the same shape.
Lindy publishes its phone pricing in credits per minute. Per the Lindy academy page, the standard configuration for incoming and outgoing US calls is 20 credits per minute. Credits run about $10 per 1,000, which is roughly $0.20 per minute, and Lindy's own blog calls it $0.19 per minute on GPT-4o. Each phone number adds $10 per month. International rates climb fast: 53 credits/min for a German mobile, 1,019 credits/min for satellite calls. The model also notes that once a Lindy phone agent forwards a call to a human, LLM charging stops because two humans are now on the line.
None of this is a knock on Lindy. It is a real product with real users, and for the workflows it was built for (CRM-shaped voice automation across general business tools) the architecture makes sense. The point of this page is narrower: a restaurant's published ordering line is not a CRM workflow. The output of every successful call has to be an order on the POS, and that is the part Gaia's integration directory does not address.
From aiphoneordering.com/llms.txt, Features section
Direct POS integration. Orders flow directly into Clover, Square, Toast, NCR Aloha, and Revel. 50+ POS integrations available. No manual re-entry.
Lindy's connector directory lists CRMs, productivity tools, and team chat. None of the five POS systems above appear in Lindy's phone-agent marketing as supported endpoints. That is consistent with Gaia's product shape, not a gap in Lindy's positioning. It is the gap that matters for a restaurant.
Same call, two outputs
Take one caller, one menu, one Friday night. The caller orders a large pizza, half pepperoni and half mushroom, gluten-free crust, light cheese, pickup in 30 minutes. On a Lindy-style horizontal agent, the call ends with a HubSpot note plus a Slack ping to the manager. On a restaurant vertical agent, the call ends with an order on the kitchen display.
Same caller, different artifacts
# HubSpot Contact Activity (Lindy/Gaia)
# Created via Lindy 'Sub-agent: HubSpot' step
# 1.4 credits used by sub-agent.
contactPhone: "+1-415-555-0142"
activityType: "PHONE_CALL"
direction: "INBOUND"
durationSeconds: 287
summary: |
Caller wants a large pizza, half
pepperoni half mushroom, gluten free
crust, light cheese. Pickup in ~30
minutes. Asked about delivery (we
do not deliver). Will call back to
confirm.
nextStep: "Call back to confirm pickup"
# Slack #orders channel:
# "[Lindy] inbound order from
# +1-415-555-0142, see HubSpot."The Lindy artifact is a CRM record describing a call. The PieLine artifact is the food order itself, structured as the POS expects it. A staff member still has to read the Lindy record and re-key the order; nobody has to re-key the PieLine record because it is already on the kitchen ticket.
The directory shape, drawn out
Lindy markets 5,000+ integrations across CRMs, productivity, and chat. PieLine has fewer connectors total, but every one of them is a restaurant POS or a restaurant-adjacent system. That is the difference between a horizontal agent and a vertical one in one diagram.
PieLine: phone call → modifier tree → restaurant POS
What Lindy's directory actually contains for a phone agent
From Lindy's own phone-agent marketing page, the integrations called out by name are CRMs, productivity tools, and team chat. Useful for a sales or support workflow. Not useful for a kitchen ticket.
None of these accept a restaurant order with item GUIDs and modifier groups.
Lindy phone-agent connectors named in the marketing copy
HubSpot
CRM contact activity log; phone calls become contact notes and tasks.
Salesforce
Enterprise CRM; Lindy can sync call transcripts to a lead or contact record.
Google Sheets
Tabular log of calls, often the cheapest way to write call summaries somewhere.
Google Calendar
Used when the phone agent is scheduling a meeting or appointment.
Google Docs
Long-form transcript storage, often used for sales call notes.
Notion
Workspace database; Lindy writes call records as Notion pages.
Airtable
Lightweight CRM substitute; one row per call, fields filled by the agent.
Calendly
Booking link automation; Lindy can text or email a Calendly URL after a call.
Slack
Real-time team notifications; the agent posts call summaries into a channel.
Now picture this list at a pizza shop on a Friday night. A caller orders. The agent writes a Notion page, an Airtable row, and a Slack message. The kitchen still has not seen the order. None of these endpoints are designed to render a half-and-half topping selection on a kitchen display.
Lindy AI Phone Agent vs. a restaurant phone agent
Same product category on the surface. Different jobs underneath. Read each row against what your ordering line actually has to do at peak.
| Feature | Lindy (Gaia phone agent) | PieLine (restaurant phone agent) |
|---|---|---|
| Output artifact per call | CRM record, Notion page, Airtable row, or Slack message; transcript plus structured fields | Order POSTed to the restaurant POS with item GUIDs and modifier groups; kitchen sees it within seconds |
| Restaurant POS connectors | None named in phone-agent marketing; directory is CRM- and productivity-shaped | Direct: Clover, Square, Toast, NCR Aloha, Revel; 50+ more on request |
| Billing model | 20 credits/min on US calls (~$0.20/min); credits cost $10 per 1,000; $10/mo per phone number | Flat $350/mo for up to 1,000 calls; $0.50/call after that; no per-minute exposure |
| Cost on 3,000 calls/mo @ 5-min avg | ~$3,000/mo in talk time alone, before sub-agent credit usage and base subscription | $1,350/mo all-in (1,000 included + 2,000 × $0.50) |
| Modifier tree understanding (half-and-half, spice levels, subs) | Captured in transcript fields; not resolved to POS item IDs and modifier groups | Native: voice utterances resolve to item IDs plus nested modifier groups per POS |
| 86 logic on voice (out-of-stock items) | Not exposed in the phone-agent configuration surface | Native: reads the POS 86 flag before quoting an item on the call |
| Daypart menus (breakfast cutoff, lunch, dinner) | Not a horizontal agent concept; would have to be hand-built in the agent flow | Reads POS menu schedule and refuses off-schedule items on voice |
| Upsell on the same call | Possible but generic; no built-in restaurant upsell library | Native restaurant upsells; AOV up 15 to 20 percent in production |
| Languages on voice | 30+ languages with auto-detection | Multilingual on the voice runtime; same modifier-tree resolution regardless of language |
| Per-account simultaneous calls | Not published; per-minute model means concurrency is bounded by spend | 20 simultaneous calls per location; predictable for a Friday night rush |
| Onboarding | DIY in a no-code canvas; you build menu scrape, POS adapter, modifier tree yourself | Done-for-you menu scrape, POS mapping, dish description model; same-day go-live |
| Good fit for a restaurant order line | No; right tool, wrong job | Yes; restaurant vertical, POS-native, 95%+ order accuracy |
| Good fit for back-office and outbound | Yes; this is Lindy's strength (catering follow-up, no-show recovery, HR, vendor calls) | Not the target use case |
Lindy column reflects public marketing on lindy.ai/blog/ai-phone-agent and the Lindy academy phone pricing page (lindy.ai/academy-lessons/lindy-phone-pricing) as of April 2026. PieLine column reflects the public llms.txt at aiphoneordering.com/llms.txt.
The per-minute math on a real restaurant
The per-minute number sounds small in isolation. Multiply it by a real ordering line and the picture changes. These figures use Lindy's own published rate of 20 credits per minute on US calls, $10 per 1,000 credits, plus a $10 phone number, against PieLine's public flat rate.
Two caveats. First, Lindy's actual monthly bill is higher than the talk-time line because every CRM or Slack write-back consumes credits too. Second, those numbers assume 5-minute calls; longer calls (catering inquiries, edge cases) push the bill higher and faster on a per-minute model than on a per-call model.
A terminal view of what each agent emits per call
If you tailed the agent log on the same Friday night for the same caller, this is what each side would write. Lindy emits CRM-shaped artifacts and a Slack ping. PieLine emits a POS write and a kitchen display push.
Concurrency and the cost-shape gap
Two numbers separate a horizontal phone agent from a restaurant phone agent. One is the per-minute rate. The other is the per-account simultaneous-call ceiling. Lindy publishes the first and not the second. PieLine publishes both.
Lindy / Gaia
0 credits/min
Standard US inbound or outbound rate, per Lindy's academy page.
Translates to ~$0.20/min, plus $10/month per phone number, plus base subscription, plus credits consumed by sub-agent steps writing to HubSpot, Salesforce, Notion, Airtable, or Slack on each call. There is no published per-account simultaneous-call cap; concurrency is effectively bounded by your monthly spend.
PieLine
0 concurrent calls
Per location, with 95 percent plus order accuracy on the AI path.
$350 flat per month for the first 1,000 calls and $0.50 per call after. No per-minute exposure on a call that goes long. Each call's output is a closed POS order, not a CRM record. Concurrency is the relevant ceiling for a Friday night, not a monthly cap.
Where Lindy is the right pick inside a restaurant group
None of this is an argument against using Lindy. Most restaurant groups already have phone workflows where a CRM-shaped output is exactly the right shape. The four below are good fits and should keep running on Lindy if you have it. The fifth is the published ordering line, where the shape changes.
Lindy fits these. The order line breaks the pattern.
- 1
1. Outbound catering follow-up
Lindy calls a recent catering inquiry, confirms head count, and writes the result back to HubSpot or Notion. CRM output is correct here.
- 2
2. No-show recovery
Lindy calls a no-show after a reservation, captures whether they want to rebook, and creates a follow-up task. Slack ping plus a CRM note is the right artifact.
- 3
3. Back-office and HR
Vendor coordination, line-cook hiring screens, supplier callbacks. Lindy plus Slack/Google Sheets handles these well.
- 4
4. The published ordering line
Pickup, delivery, dine-in callbacks. The output has to be a POS order, not a CRM note. PieLine on this number; Lindy on the others.
“Mylapore, an 11-location South Indian chain in the Bay Area, is rolling out PieLine across every site and projects $500 additional revenue per location per day from eliminating the phone bottleneck. The phone channel closes on the POS, not in a CRM.”
aiphoneordering.com/llms.txt, April 2026
Buyer's checklist: would Lindy actually fit your order line?
Run these ten checks against how your restaurant phone line actually works. If you check more than two, a horizontal agent with a CRM-shaped directory will fight your operation, not help it.
Checks that mean a horizontal agent like Lindy is the wrong shape
- Callers expect to complete a food order on the call, not be called back later.
- Modifier complexity goes beyond a sentence: half-and-half toppings, spice levels (mild / medium / spicy / extra spicy), protein subs, allergen swaps, dietary flags.
- We need 86'd items to drop out of the agent's offered list within seconds, not after a daily sync.
- We have a daypart menu (breakfast cutoff, lunch-only items, dinner specials) and the agent must respect it on voice.
- We want to quote a real total on the call so the caller commits before hanging up.
- We take card on the phone for delivery orders.
- We take more than one phone order at the same time during peak (Friday night, lunch rush, game day).
- Average call duration is 4 minutes or longer; per-minute billing is exposed to long callers.
- The phone channel should show up as a row of closed orders in the POS, not as a queue of CRM tasks.
- We do not want to maintain our own POS adapter, modifier tree, and 86 logic on a horizontal agent platform.
The one rule that decides which agent goes on which number
Most restaurant groups end up with two or more published phone numbers. The order line. A catering inbox. Sometimes a private-event line. A back-office number for staff and vendors. The simplest rule for layering AI agents across them is to match the agent to the artifact each number is supposed to produce.
On numbers where the answer is a CRM record, a Slack ping, a Notion page, or an Airtable row, Lindy is a fine fit. The 30-language voice, the 5,000-tool directory, and the no-code canvas all earn their keep there. On the published ordering line, the answer has to be a POS order. Anything that ends in HubSpot is a missed order even if it sounded great on the call.
Restaurants that fight this rule end up paying per-minute prices for a CRM record they then re-key into Toast or Clover by hand. Restaurants that follow it run Lindy on the back-office numbers and PieLine on the order line, and stop paying staff to copy summaries into a POS.
Voice-to-POS order accuracy on supported platforms
0%+
Measured on phone orders posted server-to-server into Clover, Square, Toast, NCR Aloha, and Revel. Modifier-tree resolution, not just transcript accuracy.
Keep Lindy on outbound and back office. Put a restaurant phone agent on the order line.
PieLine writes phone orders directly into Clover, Square, Toast, NCR Aloha, and Revel, with 50+ more platforms available on request. 20 simultaneous calls per location, 95%+ order accuracy, $350/month flat up to 1,000 calls. Money-back guarantee on the first month.
Book a 15 minute demo →Trade per-minute credits for a per-call POS write
Bring a Toast, Clover, Square, NCR Aloha, or Revel merchant ID and a menu. Fifteen minutes, one of your locations configured, and a live half-pepperoni-half-mushroom order resolving the modifier tree on voice and landing on the kitchen display before the caller hangs up.
Frequently asked questions
What is Lindy AI Phone Agent, in one sentence?
Lindy is a horizontal AI assistant platform; its phone product is named Gaia and lets a Lindy agent place and answer voice calls. Gaia uses the same knowledge base and integration set as Lindy's text and web agents. It is sold as a general-purpose business phone AI for support, sales, scheduling, and lead qualification, not as a restaurant ordering platform.
How does Lindy phone billing actually work for a restaurant?
Lindy charges in credits per minute, not per call. The standard configuration for inbound and outbound US calls is 20 credits per minute, and credits cost roughly $10 per 1,000, which works out to about $0.20 per minute (Lindy's own marketing rounds it to $0.19/min on GPT-4o). Each phone number is $10 per month. A 5-minute pizza order on a Friday night costs about $1.00 in talk-time credits, before any sub-agent steps consume more credits to write to HubSpot or Notion.
Does Lindy integrate with any restaurant POS systems like Toast, Clover, Square, NCR Aloha, or Revel?
No. Lindy publicly markets 5,000+ integrations and the categories named in its phone-agent marketing are CRMs (HubSpot, Salesforce), productivity tools (Google Sheets, Calendar, Docs, Notion, Airtable, Calendly), and Slack. Toast, Clover, Square for Restaurants, NCR Aloha, and Revel are not listed as Lindy connectors. PieLine writes orders server-to-server into all five of those POS platforms, plus 50+ more on request, per the Features section of aiphoneordering.com/llms.txt.
If Lindy answers the phone at a pizza shop, where does the order end up?
Wherever the Lindy agent is configured to write. In practice that means a HubSpot or Salesforce contact note, an Airtable row, a Google Sheet, a Notion page, or a Slack message. None of those are kitchen tickets. A human still has to read the message and re-key the order into Toast, Clover, Square, NCR Aloha, or Revel before the kitchen sees it. PieLine skips that step: voice utterances resolve to POS item GUIDs and modifier groups, and the order POSTs straight to the restaurant POS while the caller is still on the line.
Lindy supports 30+ languages. Does that matter for a restaurant?
It can, but it is not the bottleneck. Most independent restaurants we onboard handle two to four languages on the phone (English, Spanish, sometimes Mandarin, Hindi, Tamil, or Punjabi depending on the cuisine). Multilingual phone is table stakes for both Lindy and PieLine. The bottleneck is what the agent does with the order in the language the caller used. A multilingual transcript that ends in Slack is not a multilingual POS write; PieLine resolves the call to the same modifier-tree payload regardless of language and lands it on the kitchen ticket.
How do the costs compare for a restaurant doing 3,000 phone orders a month?
Assume an average call duration of 5 minutes. On Lindy that is 15,000 minutes at $0.20/min, or about $3,000/month in talk time alone, plus a $10/month phone number, plus the base subscription ($19.99 Starter or $49.99 Pro), plus credits consumed by sub-agent steps that write to your CRM or Slack on each call. On PieLine the same 3,000 calls cost $350 flat for the first 1,000 calls and $0.50 per call for the next 2,000, total $1,350/month with no per-minute risk. The bigger gap is what you get for the money: a Slack notification queue versus 3,000 closed POS orders.
Could you build a restaurant phone agent on Lindy's no-code platform from scratch?
You could attempt it, but you would be building the menu scrape, the modifier-tree model, the POS adapter, the 86 logic, the daypart scheduler, the upsell prompts, and the order serialization yourself, on a horizontal voice runtime that bills per minute and has no POS connectors in its directory. None of those are simple. PieLine ships them as the product and onboards the restaurant in a day. Restaurants that try to assemble a vertical agent on a horizontal tool tend to abandon the project once they hit the third POS quirk (a Toast modifier group that has 'No' as a default option, a Clover order GUID that requires line-item discount math, a Square cart that rejects unknown variation IDs).
What is Lindy's actual sweet spot if a restaurant group already has it?
Outbound and back-office automation. Catering follow-up calls. No-show recovery on reservations. Vendor coordination calls. Hiring screens for line cooks. HR voicemail triage. These are the workflows where Gaia plus Lindy's CRM and productivity integrations actually shine, because the output is supposed to be a CRM record, a Slack message, or a Google Sheet row. They are not the workflow on the published ordering line, where the output has to be a POS order.
Does Lindy publish a per-account simultaneous-call ceiling?
No. Lindy notes that a single phone number 'supports multiple concurrent calls' but does not publish a per-account concurrency cap, because Gaia is metered on per-minute talk time, not on call slots. PieLine handles up to 20 simultaneous calls per location, which is the relevant ceiling for a Friday night pizza rush. Concurrency is what determines whether a peak-hour caller hears a hold queue or hears the agent.
Where can I check the PieLine POS list myself?
It is in the Features section of PieLine's llms.txt at https://aiphoneordering.com/llms.txt. The line reads: 'Direct POS integration. Orders flow directly into Clover, Square, Toast, NCR Aloha, and Revel. 50+ POS integrations available. No manual re-entry.' That same llms.txt file is the source of truth for the 20 simultaneous calls number and the 95 percent plus order accuracy figure. We keep it aligned with what onboarding actually supports.
We use Lindy for outbound. Do we have to remove it to add PieLine?
No. Lindy can stay on outbound calls, catering follow-ups, no-show recovery, and any back-office line where the desired output is a CRM record or a Slack message. PieLine takes over the published ordering line, where the desired output is a POS order. Two different jobs, two different agents, no conflict. What you do not want is Lindy on the order line, because a half-and-half pizza that ends in HubSpot is not a half-and-half pizza on the kitchen display.
See a real restaurant phone order land on the POS in real time
We answer a test call, build the modifier tree on voice, and you watch the ticket print on your kitchen display. No CRM record in the middle. No per-minute meter ticking on a long catering call.
Book a demoHow did this page land for you?
React to reveal totals
Comments (••)
Leave a comment to see what others are saying.Public and anonymous. No signup.