AI agents that go from a WhatsApp chat to the invoice in your ERP
Meteor serves your customers on WhatsApp, Instagram, Messenger and web chat. And it does the work nobody sees: reads your ERP, syncs your catalog with the marketplaces, quotes the shipment, collects payment and logs all of it in the CRM.
Hi 👋 I need a quote for 50 units of product A10:24
Got it. For 50 units the price is COP 1,250,000 with shipping included. Want me to draw up the quote?10:24
Yes, please10:25
Done ✅ Quote #4821 sent. Shall we confirm the order now?10:25
Type a message…
Conversion today
+34%
New lead
Andrés R.
Booked a viewing · Real estate
Qualified by AI
What is an AI agent and what is it for in a company?
An AI agent is software that understands the context of your business and does real work, not a chatbot following a rigid script. At Meteor we call them Mets, and they work in two ways. Operational Mets never talk to anyone: they query your ERP, sync price and stock with the marketplaces, quote a shipment, issue a payment link or publish content. Conversational Mets handle your customers on WhatsApp, Instagram, Messenger and web chat: they answer, qualify, quote and book. Same platform, with plans from US$99 a month.
WhatsApp SIESA Instagram Alegra Siigo Messenger Microsoft Dynamics 365 KE Keybe Shopify Mercado Libre Contífico Odoo WooCommerce Éxito Mercado Pago Wompi Homecenter Cal.com Google Business Profile Rappi Andreani WordPress Aveonline Google Ads Webflow Chilexpress TCC Gmail Skydropx Wasi Google Sheets HQ Rental Google Calendar SO Softix LobbyPMS Fudo Google Drive SNIES Meta WhatsApp SIESA Instagram Alegra Siigo Messenger Microsoft Dynamics 365 KE Keybe Shopify Mercado Libre Contífico Odoo WooCommerce Éxito Mercado Pago Wompi Homecenter Cal.com Google Business Profile Rappi Andreani WordPress Aveonline Google Ads Webflow Chilexpress TCC Gmail Skydropx Wasi Google Sheets HQ Rental Google Calendar SO Softix LobbyPMS Fudo Google Drive SNIES Meta
24/7
automated operation
42
native integrations
8
Mets ready to start
4
channels in one inbox
The problem
Growth breaks all of them at once: six processes an AI agent can automate
It is not just the chat. When volume goes up, every area of the business hits the point where doing it by hand no longer covers it, and they almost always hit it together.
Customer service
WhatsApp, Instagram and Messenger: four inboxes and no owner
WhatsApp, Instagram, Messenger and the chat on your site, each on its own. The same customer asks in two channels and gets two different answers, and whoever replies first takes the sale.
Sales
The pipeline lives in someone’s head, not in the CRM
What was discussed with each customer is not written down anywhere. Knowing how much money is in play means building a spreadsheet by hand, and the next touch depends on a rep remembering.
Finance
Every quote goes through the ERP by hand
One “let me check the price and whether we have stock” turns a one-minute answer into a four-hour one. And at month end nobody has at hand which invoices are overdue, whose they are and since when.
Store and marketplaces
One price and one stock count per channel: Mercado Libre, Éxito, Rappi and your store
Your store, Mercado Libre, Rappi, Éxito and Homecenter: five portals and the same keying five times over. Copying the price straight across eats your margin in fees, and stale stock has you selling the same unit twice.
Marketing
The content calendar depends on someone remembering
Several brands, five formats and a schedule that only exists if a person opens it. Instagram does not post on its own, so publishing on time ends up being someone with a phone at seven in the morning.
Content
The design asset is the bottleneck
The campaign is already decided and the post sits waiting on the image, the video or the voiceover. When content depends on one person, their calendar sets the schedule instead of the business.
The team
Multi AI agents: it isn't one Met, it's a team, and every area has its own
Each area of your business works with its own Met, with its own tools and its own limits. Work comes in one side of the office and leaves the other side resolved.
Starts with the contentEnds with the delivery
Content
Generates the image and the video from a description, edits an image you already have, and voices the audio note that goes into the chat.
Works with · Video · Voice note
Comes with
Nano Banana Pro
GPT Image 2
Seedance 2.5
Hailuo 02
Marketing
Publishes on Facebook and Instagram (post, photo, carousel, story and reel) and handles several brands from the same workspace.
How an AI agent works in 3 steps: from a WhatsApp message to a sale
Meteor automates the whole path: it answers, qualifies and follows up on every customer so your team only has to close.
1
Connect your channels and your systems: WhatsApp, your ERP and your accounting
You link WhatsApp, Instagram, Messenger and web chat. And on the other side your store, your ERP and the tools you already run on.
2
The Mets answer and execute
They reply instantly, check real stock and pricing, quote, book, and do the work that never passes through a conversation.
3
You close and measure
Your team only gets the opportunities worth their time, with the full history and the metrics in the CRM.
The Mets
Types of AI agents: a specialized Met for each part of your business
This is not a chatbot. Each one is a Met: some talk to your customers and some work without anyone writing to them. They are the most common starting points, not a closed catalog. A Met is configured for whatever job you need.
Operational Mets
Operational AI agents: they work without anyone writing to them
They fire on an event or on a schedule, not on a message. They are the ones moving inventory, prices, orders and receivables while your team does something else.
Marketplaces
Igor
Warehouse Manager
Publishes your catalog on Mercado Libre, Rappi, Éxito and HomeCenter, and keeps price and stock current on each one. Holds back whatever does not comply before the marketplace rejects it.
ERP and availability
Siesto
Operations Director
Answers “can I sell this?” with the real figure from the ERP: stock per warehouse, that customer’s price list, their receivables and their credit limit. It never approximates.
Quotes and receivables
Legra
Accountant
Builds quotes and orders in Alegra, checks the status of your invoices and records payments and expenses. The bookkeeping, in plain language.
Conversational Mets
Conversational AI agents: they look after your customers
They talk to whoever writes to you on WhatsApp, Instagram, Messenger and web chat, and leave every conversation on record in the CRM.
24/7 on your channels
Customer support
Answers questions, resolves issues and looks after your customers on WhatsApp, Instagram, Messenger and web chat, at any hour.
Prioritize real buyers
Lead qualification
Filters and scores every prospect automatically so your team only talks to the ones ready to buy.
Close in the chat
Closing sales
Walks the customer to the purchase: handles objections, recommends products and closes the sale inside the chat.
Appointments and bookings
Scheduling
Handles bookings and appointments on its own, in sync with your calendar and availability.
Create, publish and advertise
Content and social
Creates the image or the video, publishes to Facebook and Instagram, and builds the ad campaigns on Meta and Google. You approve before it publishes, and before it spends.
The operational side is not the extra. Much of a Met’s work happens with no conversation involved at all: syncing the catalog, reading the ERP, issuing the invoice.
Build it yourself, buy it with AI on top, or Meteor
Anyone can get the AI model. The hard part is everything around it: connecting it to WhatsApp and to your ERP, having it remember each customer, and keeping it running the day a vendor changes its API.
How it works
What sits behind each reply: which model handles it, and who runs it.
How it works: Comparison of capabilities between building an agent on top of a model, a legacy software vendor with AI bolted on, and the Mets of Meteor
Capability
You build it A model API, and you write and maintain everything else.
Software with AI on top A product built for something else, with AI as a separate module.
Meteor AI-native: the Met is the product, and the engineering is already done.
A model that understands natural language The starting point for all three options, and the advantage of none: anyone can get the model today.
Yes. It is exactly what you buy
Yes. It ships in the AI module
Yes. The base, not the product
The right model for each task, not one model for everything Holding a conversation, transcribing audio, reading a photo of a handwritten slip and generating an image are not the same task, and one model does not do all of them well. There are more than 25 connected; the system picks by task type and drops to the lightest one that will do.
Partly. You can, but you wire it
No. Whichever the vendor integrated
Yes. And it degrades on its own under load
A team of Mets, not one agent doing everything Each Met has its role and its own tools, and whoever takes the message hands the work out by skill. You can take a tool away from one without touching the others, and test it in a sandbox conversation before it serves a real customer.
No. One prompt that grows unchecked
No. An assistant, not a team
Yes. Roles, tools and a sandbox
What it connects to
Reaching your customer where they already are, and the data you answer with.
What it connects to: Comparison of capabilities between building an agent on top of a model, a legacy software vendor with AI bolted on, and the Mets of Meteor
Capability
You build it A model API, and you write and maintain everything else.
Software with AI on top A product built for something else, with AI as a separate module.
Meteor AI-native: the Met is the product, and the engineering is already done.
Official WhatsApp Business API, Instagram and Messenger connection This is not plugging in an API: it is number verification, the business account and the rules of each channel.
No. You file for verification yourself
Yes. Usually comes solved
Yes. Connected and verified
The same Met that talks reads stock, pricing and receivables from the ERP No copy-pasting between screens: the figure is read by the same one doing the answering, at the moment it is needed.
No. One connector per system, yours
Partly. The product does; the AI, separately
Yes. It is the same conversation
What is left afterwards
What outlives the conversation: the record, the log, and who keeps it running.
What is left afterwards: Comparison of capabilities between building an agent on top of a model, a legacy software vendor with AI bolted on, and the Mets of Meteor
Capability
You build it A model API, and you write and maintain everything else.
Software with AI on top A product built for something else, with AI as a separate module.
Meteor AI-native: the Met is the product, and the engineering is already done.
Remembers what it already did for that customer and fills in their record It arrives at the conversation with the last tools it ran for that contact, so it does not requote what it already quoted. On close, it fills the empty CRM fields and scores intent, never overwriting what a person wrote.
No. Every call starts from zero
Partly. Stores the history; does not use it
Yes. Memory and capture, out of the box
You see what each Met did, what it cost and whether it went well Every run is logged with its Met, its type and its cost, along with the success rate and service quality of each one. And in the chat you can open the reasoning: which tool it called, what it called it with and what came back.
Partly. Only what you instrument
No. Measures the license, not the agent
Yes. Per Met and per run
You do not run servers, or retries, or the day a third party changes its API Backoff retries, signing every webhook, discarding duplicate messages and renewing expiring tokens all run on their own. And when the channel or the ERP changes its API, we are the ones who find out and fix it.
No. You write it and you watch it
Yes. Also their job
Yes. Included, without asking
Meteor’s column only marks “yes” where the capability is built and running, and it describes only what you can see from your own account. The other two describe what each path is by definition: starting from a model API gives you the model and nothing else; legacy vendors bring the product and the AI separately. It is not an assessment of any specific vendor.
The platform
AI agent platform with CRM and automation to serve, sell and execute
AI, CRM, automation and integrations in one platform. Every piece removes manual work: answering, and moving data from one system to another.
Conversational AI
Agents that understand, answer and sell in natural language on WhatsApp, Instagram, Messenger and web chat, in your brand’s voice.
Sales CRM
Every customer, order and quote in one place, with the full history of the conversation and the deal.
Workflow automation
You design the whole process on a visual canvas, with the Met inside it: more than 35 node types, including conditions, loops and HTTP calls to any API. The Met reasons; the flow repeats the same steps every time, and that is where the things that cannot fail go: charging, dispatching, notifying.
Catalog and quotes
Show products, build quotes and close orders inside the chat, without pulling the customer out of the conversation.
Lead follow-up
No prospect goes cold: reminders, automated sequences and smart reassignment of opportunities.
Reports and metrics
Real-time visibility into conversations, conversion, response times and team performance.
SDK, CLI and MCP server
Point Claude, Cursor or Codex at your operation: the MCP server exposes 34 tools (run a Met, move the CRM, fire a flow), each limited to the scopes of the key. With a TypeScript SDK, a Python SDK and a CLI.
Answers stock and price for the exact variant, builds the order and takes payment inside the chat. And publishes your catalog on Mercado Libre, Éxito, HomeCenter and Rappi with price and stock current.
Create the piece, publish to Facebook and Instagram, and build the campaigns on Meta and Google, including the ads that open a WhatsApp chat on your number. Nothing starts spending until you turn it on.
Native integrations: AI agents connected to your ERP, your accounting and your channels
A Met does not live inside the chat. It greets your customers in one inbox with one history, queries the system where the data lives (real stock, the customer’s price, receivables), syncs price and stock where you already sell without opening each panel, quotes the shipment, issues the payment link, publishes on its own, books the appointment and files the customer in the same message, and answers with the figure from the source. These are the native integrations, and there is a REST API for whatever is missing.
Legra
Alegra
Issues your electronic invoices and tells you where your receivables stand.
Siesto
SIESA
Checks stock, prices and order status in your ERP.
Meli
Mercado Libre
Moves your catalogue and answers the marketplace questions.
Igor
Bodegas
Keeps your store stock in sync with the marketplaces.
ERP, accounting, marketplaces, logistics and payments: what is already connected
42 native integrations, with their type in plain sight, plus the four image and video models that come built in.
42 native integrations and counting, plus the image and video models that come built in. Your system not on the list? It connects through Meteor’s REST API.
Onboarding
How to build an AI agent for your company: 8 steps, guided from start to finish
We do not hand you a tool and wish you luck. We walk you through it until Meteor runs on its own.
01
Discovery
We map your operation, current channels, message volume and commercial goals.
02
Channel setup
We connect WhatsApp, Instagram and wherever else your customers are.
03
Flow design
We build the sales, support and follow-up flows around how your business actually works.
04
AI training
We feed the agents your catalog, your brand voice and your key answers.
05
Integrations
We wire Meteor into your e-commerce, CRM, logistics and internal tools.
06
Testing
We validate every flow against real scenarios before your customers see it.
07
Launch
We go live with the Meteor team monitoring alongside you.
08
Continuous tuning
We measure, adjust and improve the results month over month, on data.
Benefits
What AI agents are for: less manual work, more sales closed
24/7
Support that never pauses
Your business answers and sells outside business hours too.
Seconds
Instant first reply
Every customer gets a first answer right away, at any hour.
0
Leads without follow-up
Reminders and automated sequences so no prospect goes cold.
100%
Conversations on record
Every chat, order and quote stored and measurable in the CRM.
1
One operation, one place
Customers, orders and quotes in a single place, not across five tools.
42
Integrations ready to go
Connect your store, your logistics and your tools with no custom build.
Testimonials
Companies already using AI agents: real businesses selling with Meteor
Do not take our word for it: hear it from the ones who already automated.
Every plan includes the Ágil, Brillante and Genio AI models. AI usage is billed as Energy based on what you actually use, and every plan includes an Energy allowance in every period you pay for. Pay annually and two months are free.
What an implementation includes
What it takes to get AI agents running with Meteor
A clear base scope to launch with results, and add-ons for when you need to scale.
Recommended to start
Base implementation
Everything needed to get Meteor operating inside your business.
Discovery and map of your commercial operation
WhatsApp setup plus one additional channel
Up to 3 sales, support or follow-up flows
AI training on your catalog and brand voice
Conversational CRM configured
1 integration (e-commerce, logistics or API)
Testing, launch and team training
30 days of tuning and hands-on support
Add-ons and scale
Add capabilities as your operation grows.
Additional channels (Instagram, Messenger, web chat)
Advanced and multi-team flows
Custom integrations through the API
Automated catalog and quoting
Advanced reporting and custom dashboards
Ongoing monthly support
Each implementation is quoted on channels, volume and level of automation. This proposal sets the scope; the figures are adjusted to your operation.
Meet the team behind Meteor.
FAQ
Frequently asked questions about AI agents: what companies ask before they start
Is Meteor a chatbot?
No. A chatbot follows a rigid script; a Meteor Met understands what the customer writes, decides what to answer and moves the sale forward: it quotes, books, recommends and closes inside the conversation. And it also does work outside the chat, against your ERP, your marketplaces and your carrier.
Does Meteor have a free trial?
There is no free plan for running the platform: Meteor plans start at US$99 a month. There is a US$0 Developer plan for API, SDK and MCP server access, and it is not for sale: it is granted when you are approved as a Tech Partner or a Startup Partner. Before you buy, you can ask for a live demo at no cost.
How much does Meteor cost?
Meteor plans run from US$99 a month (Estudio) to US$3,000 a month (Campus), depending on the size of your team and your contact volume. Pay annually and two months are free. AI usage is billed separately as Energy, and every plan includes an Energy allowance in every period you pay for, from US$10 to US$450 depending on the plan.
Does it work with the WhatsApp Business API and Instagram?
Yes. Meteor runs on WhatsApp, Instagram, Messenger and web chat, and centralizes every conversation in one CRM. WhatsApp is the most used channel, but it is neither the only one nor the product.
Do I need the WhatsApp Business API already set up to use Meteor?
You do not need to have it sorted out beforehand. The Meteor team walks you through connecting your WhatsApp line during onboarding, along with the rest of your channels. If you already have a line running, it connects to Meteor without changing the number.
Are Mets only good for conversations?
No. A conversational Met handles your customers across channels, and an operational Met does work that never passes through a chat: it syncs price and stock with Mercado Libre, Éxito, HomeCenter or Rappi, reads inventory and receivables from your ERP, quotes a shipment with the carrier, issues a payment link or publishes to your social accounts. It is the same platform.
Does it integrate with my store and my tools?
Yes. Meteor has 42 native integrations: stores (Shopify, WooCommerce), marketplaces (Mercado Libre, Éxito, HomeCenter, Rappi), ERP (SIESA, Microsoft Dynamics 365), accounting (Alegra, Siigo), shipping (Aveonline, TCC, Skydropx), payments (Mercado Pago, plus Wompi for PSE and Nequi in Colombia), websites (WordPress, Webflow) and calendars. For whatever is not on the list there is a REST API, an SDK and an MCP server.
How much work can a Met actually get done?
Across the whole Meteor platform, more than 18,000 tasks run every month. A task is a complete piece of work: answering a customer, syncing a product with a marketplace, quoting a shipment, issuing a payment link or logging the conversation in the CRM. Mets work with no shift pattern, so the volume does not depend on anyone being on the clock.
Which countries does Meteor work in?
Meteor’s main market is Colombia, and it also operates in Mexico, Argentina, Ecuador, Curaçao, United States, Chile and Nicaragua. The channels and the CRM work in any country; what changes by market are the local integrations, such as Alegra for electronic invoicing with the DIAN in Colombia, or Mercado Pago and Mercado Libre in Mexico and Argentina. Plans are billed in US dollars.
How long does implementation take?
Weeks, not months. The Meteor team walks you through it step by step: discovery, channel setup, flow design, AI training, integrations, testing and launch.
Do I need a technical team to use Meteor?
No. The Meteor team configures the flows and trains the Mets on your catalog and your brand voice. You focus on selling. And if you do have a technical team, you can build on the REST API, the SDK and the MCP server.
What happens to my customers’ data?
The data stays yours. The data controller is Rocketfy, Inc., the company behind Meteor, declared with its name, address and privacy contact in the site’s privacy policy. You can request deletion of your data at any time from the data deletion page or by writing to privacy@meteor.com.co.
Who is behind Meteor?
Meteor is the product of Rocketfy, Inc. Colombia is its main market, and it is where Meteor has the most verified local integrations. It has no relation to the JavaScript framework of the same name.
Start today
Turn every conversation into a sale
Book a demo and we will show you, on your own case, how Meteor automates your operation in weeks: from the first message to the system where the data lives.