My goal is to replace the repetitive manual work inside my SMM panel with an AI-driven agent that talks fluently to both the panel’s API and WhatsApp. Core functions for the panel • Order management – every time a client creates a support ticket the agent must read the ticket, extract the order ID, then call the provider’s API endpoint and pass that ID along with the requested action. • Cancellation handling – if a WhatsApp user writes “cancel” together with a valid order ID, the agent should instantly push a cancellation request through the same API. • Service announcement – whenever I add a new service the agent has to broadcast a short promotional message to every group I belong to inside the panel, up to 30 times per day (roughly once an hour). WhatsApp side The bot needs to understand three message intents: order placement, order cancellation and order status inquiry. Free-form text is fine as long as the user includes an order ID; the agent must parse it, map it to the right action, hit the panel API and reply back with a friendly confirmation. Integration details Everything has to run through the SMM panel’s official API. If the API ever returns an error the agent should retry with exponential back-off and log the response for later review. Preferred tech Python, Node.js or anything equally stable is acceptable. Using OpenAI or similar LLMs for intent detection is fine, but I also welcome rule-based NLP if it keeps latency low. For WhatsApp connectivity you can rely on Twilio, Meta Cloud API or a comparable provider. Deliverables 1. Source code with clear README explaining environment variables, API keys and deployment steps 2. Tested endpoints for: create order, cancel order, check status, push new-service broadcast 3. A simple dashboard or log file that records every action, response code and timestamp 4. One-hour live walkthrough so I can see the agent working inside my own panel and WhatsApp number Acceptance criteria The agent must successfully handle at least 20 live panel tickets, 20 WhatsApp messages and 3 new-service broadcasts in a row without human intervention. If you can build, test and deliver this, let’s talk.