Build a Custom
AI Agent

A custom AI agent on ewpire is a purpose-built AI employee you configure around your specific business context – your brand voice, your customers, your knowledge base, and the tools you need. You start from one of 9 expert-tuned templates (Sales, Support, Research, Lead Qualification, Growth, HR Screener, Tender, Document Processing, or Executive Concierge), then layer your own instructions on top via a guided chat in Telegram. The result is an AI employee that operates 24/7 inside your existing workflows, using your company data and following your escalation rules – not a generic AI assistant, but a specialist configured to your exact standards.

No. Building an AI agent from scratch means writing prompts from zero, wiring tools, handling edge cases, and iterating for weeks before you see reliable output. On ewpire, you start from a template that ships with battle-tested logic: the Sales Agent already knows 4 email frameworks and trigger-event detection. The Tender Agent already understands 18-field bid analysis. Your job is to customize – not build. You supply your company name, target ICP, brand voice, and knowledge base. The underlying expert behavior is already there. Think of it like hiring a specialist and briefing them, rather than training one from zero.

No code required. Configuration happens entirely through the /onboard command in Telegram – a conversational 7-step wizard that walks you through brand voice, ICP definition, content languages, tool preferences, and escalation rules. Upload knowledge base files via drag-and-drop. Switch active agents with /use, adjust settings with /settings, and reset memory with /reset. If you can send a message, you can configure and manage your AI employee.

Initial setup takes around 10 minutes: pick a template, run /onboard, answer the guided questions, upload your key docs. The agent is live after that. The first week is when the real tuning happens – you interact, spot where the output isn't quite right, and adjust via /settings or /reset. Most teams find the agent hits a stable, high-quality output rhythm within 5–7 days of actual use. There is no fixed "training period" – changes take effect on the next message.

Every configurable dimension of the agent is tunable: the system prompt and role description, brand voice (formal, friendly, technical, bold, or playful), your ICP and target customer definition, content output languages from 99+ options, which tools are enabled (email send, CRM lookup, calendar, web search), escalation rules and approval mode, and agent-specific knobs – for example the Sales Agent's email framework priority, the Tender Agent's sector focus, or the HR Screener's must-have skills. Upload a knowledge base of PDFs, URLs, FAQs, and SOPs that the agent retrieves from when answering. Business plan adds custom brand voice training on your own content corpus.

Yes. During onboarding or any time after via /knowledge, you can upload PDFs, paste URLs, or provide plain-text FAQs and SOPs. The agent ingests, chunks, and indexes everything using vector search so it retrieves the most relevant context for each query – and cites the source document in its answer. Storage limits by plan: Starter 10 MB, Pro 500 MB, Business 5 GB. The agent only answers from your uploaded knowledge – it does not fabricate information. Add new documents any time and the index updates immediately.

Lindy and Relevance AI are no-code workflow builders – you start from a blank canvas and assemble an agent piece by piece. That is powerful for unusual use cases, but it takes time and domain knowledge to reach production quality. ewpire ships 9 expert-tuned templates out of the box, so your Sales Agent already understands email frameworks, and your Tender Agent already knows bid analysis without you building it. Our agents also run natively inside messengers your team already uses – not inside a separate web dashboard you have to train users on. On Pro and Business, BYOK lets you bring your own Claude, GPT, or Gemini API key if you prefer a specific model.

Yes. Pro plan lets you add up to 3 team operators via /invite. Business plan supports up to 10. Each operator connects to the shared agent through the same Telegram bot – they send messages, run commands, and receive reports directly in their own chat. The agent enforces exclusive active-user locking so two operators cannot send conflicting instructions simultaneously. Administrators see per-operator usage stats and can revoke access at any time. The shared agent memory persists across operators, so context carries between team members.

The 9 templates cover the workflows that generate the most repetitive overhead in text-heavy businesses: outbound sales, customer support, competitive research, lead qualification, content and marketing, HR screening, tender management, document processing, and executive task management. These apply across professional services, B2B services, consulting, legal, healthcare-adjacent, finance, e-commerce, logistics, manufacturing, and more. If your core workflow is one of those categories, you configure the template to your industry context – sector-specific vocabulary, compliance language, your document types. If your workflow is genuinely novel and does not fit any of the 9, use the request form at the bottom of this page.

You iterate. /reset clears the agent's conversation memory so you start fresh without losing your configured settings. /settings lets you edit brand voice, ICP, or any parameter mid-stream. /transfer moves active context between agents if the task turns out to belong to a different specialist. Nothing is permanent – every change takes effect immediately. And if after real use the agent is genuinely not working for your use case, our 14-day money-back guarantee applies across all plans. Email [email protected] within 14 days of first charge.