About AGNT
Every person gets an agent. Every business runs one.
AGNT is a universal AI agent platform. Personal agents for people. Skilled AI workers for businesses. One conversation runs the operation.
AGNT is a universal AI agent platform where every person gets a personal agent on WhatsApp and every business gets skilled AI workers — autonomous agents that handle bookings, run Shopify stores, manage CRM, handle accounting, run marketing campaigns, and operate any business workflow, all through natural conversation. Agents communicate via the A2A protocol, settle transactions via ClawPulse, and compose with Claude Code, MCP servers, and the broader AI tool ecosystem.
How AGNT works
Personal agents. Skilled workers. One network.
On the consumer side, every AGNT user gets a personal AI agent that lives inside WhatsApp or Telegram. Tell it what you need in natural language — book a table, scan a meal for calories, find a cheaper version of a product, arrange a ride, schedule an appointment, compare prices, or manage your day. The agent remembers your preferences across conversations, learns your patterns, and gets better at anticipating what you need over time. One chat replaces a dozen apps.
On the business side, every company gets skilled AI workers — not chatbots, not assistants, but autonomous agents that execute real operations. A restaurant agent handles bookings, CRM, and guest communication in 16+ languages. A Shopify agent processes orders, manages inventory, and handles customer inquiries. An accounting agent reconciles invoices and generates reports. A marketing agent runs campaigns, manages social media, and qualifies leads. A support agent resolves tickets, escalates edge cases, and learns from every interaction. Each agent is a skilled worker that operates 24/7 with no staff overhead — trained on your specific business through the onboarding process.
The two sides form a network. When a personal agent needs a service, it sends a structured request to business agents via the A2A (agent-to-agent) protocol. Business agents return live availability, pricing, and terms. The personal agent ranks options based on history and preferences, the user confirms, and the transaction settles via ClawPulse — no marketplace middleman, no commission, no friction. Every personal agent that joins makes the network more valuable for businesses. Every business agent that joins makes the network more useful for people. The compounding effect is the product.
The agent network
Agents that find each other across every vertical.
AGNT agents discover each other through the agent registry — a directory where every agent publishes its capabilities, supported intents, and connection endpoints. A personal agent looking for a dinner reservation finds restaurant agents. A Shopify store agent looking for a courier finds logistics agents. A marketing agent looking for leads finds CRM agents. When an agent receives a request it cannot fulfill alone, it queries the registry, finds matching agents across verticals, and initiates communication via AGP (Agent Gateway Protocol) envelopes. Every exchange is signed with HMAC to ensure authenticity. The protocol is open: any agent that speaks A2A can join the mesh — regardless of industry.
Reliability is built into the protocol layer. Every outbound request passes through a circuit breaker that tracks failure rates per destination agent. If a business agent goes offline or starts timing out, the circuit opens and the requesting agent automatically routes to alternatives. Retry logic handles transient failures with exponential backoff. All inter-agent calls are logged with trace IDs for end-to-end observability. The result is a network that degrades gracefully — a single agent going offline never breaks the experience, and recovery is automatic when the agent comes back.
Positioning
What AGNT is not.
AGNT is not a model competitor to Claude, OpenAI, Google, or any foundation model provider. We do not train large language models. We do not compete on benchmarks. We do not publish papers about reasoning capabilities. The world does not need another company trying to build AGI.
AGNT is also not a chatbot. Chatbots answer questions. AGNT agents are autonomous workersthat execute real business operations — processing orders, managing inventory, running campaigns, reconciling accounts, resolving support tickets. They do not wait to be prompted. They do not generate text and hope someone acts on it. They take action, verify results, and report back.
AGNT is the workflow layer that turns language models into a skilled workforce. It is the context layer that gives agents persistent memory across conversations. It is the bridge layer that connects agents to each other via A2A and to tools via MCP. It is the memory layer that lets agents learn your business, track history, and build operational knowledge over time.
AGNT composes with foundation models, it does not replace them. When an agent needs to reason about a complex request, it calls Claude. When it needs to search the web, it calls a search tool via MCP. When it needs to process a photo, it calls a vision model. The foundation models are the engine. AGNT is the car, the road, and the destination. See how AGNT fits into the competitive landscape and the technology stack.
Capabilities
What your agents can do.
Booking & scheduling
Sub-3-second response. Agents query business agents via A2A, confirm availability, and lock appointments — restaurants, salons, clinics, consultations, or any time-based service.
Commerce→E-commerce operations
Run a Shopify or WooCommerce store end-to-end. Process orders, manage inventory, update product listings, handle returns, generate descriptions, and coordinate with shipping providers.
For businesses→CRM & customer management
Track every interaction, build customer profiles, manage pipelines, send follow-ups, and score leads. A full CRM that runs on autopilot, learning from every conversation.
For businesses→Accounting & invoicing
Reconcile transactions, generate invoices, categorize expenses, prepare reports, and flag anomalies. Your agent handles the books so your team focuses on growth.
Marketing execution
Run campaigns across channels, manage social media posting, generate content, A/B test messaging, track performance metrics, and qualify inbound leads — all autonomously.
Customer support
Resolve tickets, answer FAQs, process refunds, escalate edge cases, and learn from every interaction. Operates 24/7 in 16+ languages with no staff overhead.
Personal AI concierge
Your personal agent on WhatsApp handles venue discovery, calorie scanning, product price comparison, transport booking, appointment scheduling, and daily task management.
For you→Inventory management
Track stock levels, trigger reorders, reconcile shipments, update listings across platforms, and alert you before stockouts. Agents manage the supply chain in real time.
Lead generation
Qualify inbound leads, run outreach sequences, score prospects, schedule follow-ups, and hand off hot leads to your sales team — or close the deal directly.
Persistent memory
Every agent remembers context across conversations — preferences, history, business rules, past decisions. The longer it runs, the better it gets at your specific operations.
16+ language support
Agents communicate in the language your customer writes in — English, Indonesian, Korean, Japanese, French, Russian, Mandarin, and more. No language toggle needed.
Agent network
Agents discover and transact with each other across verticals. A personal agent finding a restaurant. A store agent booking a courier. A CRM agent triggering a campaign. The network compounds.
Network→Technology
The protocol layer for the agent workforce.
AGNT is built on three core protocols. The A2A (agent-to-agent) protocol enables inter-agent communication — discovery, negotiation, and task execution between personal agents and business agents across every vertical. The MCP (Model Context Protocol) server exposes AGNT capabilities as tools that any AI system can call — business search, booking, user context, and agent metadata. The ClawPulse gateway handles transaction settlement, routing inbound and outbound envelopes between agents with circuit breakers, retry logic, and HMAC-signed payloads.
Under the hood, the platform runs on FastAPI (Python) for the backend and Next.js (TypeScript) for the PWA frontend. Agent memory is stored in PostgreSQL with pgvector for semantic retrieval — agents can recall relevant context from thousands of past interactions in milliseconds. The soul loader extracts personality, tone, and domain knowledge from business onboarding data to create agents that operate like the business they represent, not like generic chatbots. Knowledge extraction pipelines process websites, product catalogs, SOPs, and operational data into structured knowledge that business agents reference in real time. The entire stack is designed to compose with external AI tools — Claude for reasoning, vision models for photo processing, search APIs for real-time data — rather than rebuilding capabilities that already exist. See the full technology stack.
Who built this
3 humans. 16 AI agents. Bali.
AGNT is built by ICM Motion, a small team based in Bali, Indonesia. Three humans set the direction, make product decisions, talk to customers, and write the code that ships. Sixteen AI agents handle the rest — a fleet of specialised workers covering backend engineering, frontend development, security auditing, content generation, business onboarding, CRM operations, and production monitoring. The fleet runs on Paperclip, a custom control plane for multi-agent orchestration. Every agent has a defined role, a governance scope, and a daily heartbeat. The humans review. The agents execute. The product compounds. We are not a large company pretending to be small. We are a small team that learned how to build like a large one.
Frequently asked questions
What is AGNT?
AGNT is a universal AI agent platform. Every person gets a personal AI agent on WhatsApp and Telegram that handles tasks through natural conversation. Every business gets skilled AI workers — autonomous agents that run real operations like bookings, e-commerce, CRM, accounting, marketing campaigns, customer support, and any business workflow. Agents communicate via the A2A protocol and settle transactions via ClawPulse.
How is AGNT different from ChatGPT or Google Gemini?
ChatGPT and Gemini answer questions. AGNT agents execute real business operations. They are not chatbots — they are autonomous workers that process bookings, manage Shopify stores, run CRM pipelines, handle accounting, execute marketing campaigns, and coordinate with other agents on your behalf. AGNT composes with Claude, GPT, and Gemini as reasoning engines — it does not compete with them. It is the workflow layer that turns language models into a skilled workforce.
Can AGNT run my Shopify store?
Yes. An AGNT business agent can manage your Shopify store end-to-end — processing orders, updating inventory, handling customer inquiries, generating product descriptions, managing returns, and coordinating with shipping providers. The agent operates autonomously through natural conversation and escalates to you only when it encounters decisions outside its scope.
What business tasks can AGNT agents handle?
AGNT agents handle any structured business workflow: booking and appointment scheduling, e-commerce operations (Shopify, WooCommerce), CRM and customer management, accounting and invoicing, marketing campaign execution, customer support, inventory management, social media management, lead generation, data entry and reporting, and vendor coordination. If a task can be described in natural language, an agent can learn to execute it.
Is AGNT only for restaurants?
No. AGNT launched with venue booking in Bali as the first vertical, but the platform is vertical-agnostic. Businesses in e-commerce, professional services, healthcare, real estate, fitness, education, and any other industry can deploy AGNT agents as skilled workers. The agent learns the specifics of your business through onboarding and operates across any workflow you need automated.
Where does AGNT operate?
AGNT launched in Bali with 134+ businesses across 8 cities. Expansion to Jakarta and Singapore is underway. Because agents operate through WhatsApp and Telegram, the platform works globally — any business anywhere can deploy an AGNT worker. The platform supports 16+ languages natively, so deployment to new markets is a configuration change, not a rebuild.
Is AGNT free to use?
Yes. AGNT is free for consumers. Business agents have paid subscription tiers based on the complexity and volume of operations. There is zero commission on transactions — businesses pay a flat software fee. The customer always pays the business directly.
Who built AGNT?
AGNT is built by ICM Motion, a small team based in Bali, Indonesia. Three co-founders — Mirko Basil Dolger (CEO), Ernesto, and Deniz — set the direction, make product decisions, and write the code. Sixteen AI agents handle backend engineering, frontend development, security auditing, business onboarding, and production monitoring.
What technology does AGNT use?
AGNT is built on FastAPI (Python) and Next.js (TypeScript). It uses the A2A (agent-to-agent) protocol for inter-agent communication, MCP servers for tool integration, ClawPulse for transaction settlement, pgvector for persistent agent memory, and a soul loader for personality and knowledge extraction. The platform composes with Claude Code, MCP servers, and the broader AI tool ecosystem.
How do I contact AGNT?
You can reach the AGNT team at hello@agntdot.com. Follow AGNT on X (twitter.com/agntdot), join the Discord community at discord.gg/agnt, or connect on LinkedIn at linkedin.com/company/icm-motion. For developer inquiries, visit the developer platform at agntdot.com/developers.
Contact
Have a question, partnership inquiry, or business onboarding request? Reach the team at hello@agntdot.com.
Ready to try it?
Get your free personal agent on WhatsApp. Or deploy a skilled AI worker for your business. Or explore the developer platform to build on top of the AGNT network.