Missed call = lost customer
Nights, weekends, lunch breaks — calls go unanswered and prospects move on to your competitor.
Sekretar24 is a virtual AI assistant for your business. It replies to customers on Telegram, WhatsApp, your website and over the phone, works from your knowledge base and pushes hot leads straight into your CRM.
No credit card · Full-feature trial · Cancel in one click
Nights, weekends, lunch breaks — calls go unanswered and prospects move on to your competitor.
By the time they get back, the customer has cooled off or already bought elsewhere. Reply speed wins deals.
The same FAQs, manual qualification, copy-pasting into CRM — instead of selling and growing.
Sekretar24 answers instantly and 24/7 across every channel, qualifies the lead and creates a deal in your CRM — so your team can focus on what matters.
AI Employees
From a single platform you can build an AI employee for any job — with its own prompt, knowledge base, voice and integrations. Proven scenarios:
Inbound and outbound calls, speech recognition, voice replies, lead qualification, smooth handoff of complex cases to a human.
24/7 L1 line on Telegram, WhatsApp and your website chat. Answers from your knowledge base, ticketing, escalation to engineers, quality analytics.
Phone reception, booking meetings into Google Calendar, spam filtering, reminders, conversation and email summaries.
Legal advice grounded in an up-to-date base (codes and case law for RU and KZ), contract review, document templates.
Answers on taxes and reporting, USN/VAT calculations, deadline reminders, first-pass document processing.
Sales funnel on Telegram, WhatsApp and your widget, qualification, upsells, accepting payments, amoCRM integration.
AI briefs, generating spec docs, stack selection, landing-page prototyping, ongoing client-site support.
Bridge to Claude Code: writes and reviews code, works with git, runs tests and deploys — right from chat.
Need a role that isn't on the list? Get in touch — we'll build one for your case.
How it works
Create an account — your 14-day free trial starts immediately.
Pick a ready-made role, upload your knowledge base, set the prompt and voice.
Telegram, WhatsApp, web widget and telephony — all in a few clicks.
Your assistant replies around the clock; hot leads fly straight into your CRM.
Features
By default runs on Claude by Anthropic — the smartest and most steerable model on the market. On request: GPT-4, Gemini, DeepSeek, Qwen, Llama or your own local model.
Upload documents, websites or RSS — your assistant answers strictly from your data, not hallucinations.
Speech synthesis and voice cloning — your assistant sounds like a real member of your team.
Run on cloud models or deploy the AI on your own server — data never leaves your perimeter.
amoCRM, WooCommerce, Google Calendar, Drive and Gmail — your assistant lives inside your tools.
Russian, English and Kazakh — your assistant speaks to customers in their language.
LoRA fine-tuning on your data — the assistant picks up your company's tone and expertise.
Open-source · MIT
All Sekretar24 code is published on GitHub under the MIT license — download it, deploy on your own server and use it for free. Any Linux server with Docker and root SSH access is enough.
Open source, MIT license. Free forever. Modify for your niche.
Open on GitHubRepo wiki: installation, deployment, LLM setup, RAG, telephony, bots and integrations.
Open WikiWe deploy on your server, configure the knowledge base, connect channels, tune the assistant for you. All you need is Docker and root SSH.
Order installation🔒 Self-hosted means customer data stays on your server. Run on free OpenRouter models or plug in your own Claude/GPT-4 key.
Live example
The assistant grasps the question, asks clarifying details and drives the customer to the next action — booking, request or payment. No "command not recognized", no menu trees.
Try it with your own questions📱 Android app
Native Android app. Voice chat with your assistant, push notifications for new messages and app updates, access to all your chats — from anywhere.
💡 On iPhone our mobile browser version (PWA) works fully — all assistant features are available without installation. Native iOS app is on the roadmap.
Pricing
From freelancer to agency: cloud bots and widgets, local AI and on-premise — for any budget.
For freelancers & bloggers
billed monthly
Try freeFull cloud: bots and widgets
billed monthly
Try freeLocal AI and on-premise
Pricing = setup + ongoing support
Talk to salesAfter the trial
The 14-day free trial runs on Claude by Anthropic — the market flagship. After the trial you pick one of two paths:
Business+ subscription — max answer quality, best reasoning, long context.
Switch to OpenRouter: DeepSeek R1, Llama 3.3 70B, Qwen 2.5 72B and more. The chain is configured from strongest to weakest — if the top model hits a rate limit, the assistant automatically falls back to the next one. No downtime.
💡 Free models cover most business tasks. Paying for the top tier makes sense only when you need complex reasoning and maximum expertise.
Questions
Regular bots run rigid scenario trees. Sekretar24 is real AI powered by Claude (Anthropic): it understands free text, remembers conversation context and answers from your knowledge base, not from a script.
By default — Claude by Anthropic, the smartest and most steerable model on the market. On request we plug in any other: OpenAI GPT-4, Google Gemini, Moonshot Kimi, DeepSeek, OpenRouter (dozens of models in one API) or local Qwen/Llama/DeepSeek via vLLM on your own server.
The trial runs on Claude. After — two options: stay on Claude with Business+, or switch for free to OpenRouter's powerful models (DeepSeek R1, Llama 3.3 70B, Qwen 2.5 72B). The chain is configured from strongest to weakest: if the top model hits a limit, the assistant automatically takes the next one. For most tasks free models are more than enough.
Yes. Code is open on GitHub under MIT, with detailed documentation. You need Docker and root SSH access on any Linux server. If you'd rather not bother — we install turnkey for 5,000 ₽.
For most business tasks OpenRouter's free models (DeepSeek, Llama, Qwen) are enough. Paying for top models (Claude, GPT-4) makes sense for complex reasoning and maximum expertise. On cloud plans the API is included; on self-hosted you plug in your own key or use the free models.
We deploy Sekretar24 on your server (Docker + root SSH required), configure the knowledge base (RAG), connect Telegram/WhatsApp/widget/telephony, and tune the assistant for your niche. Payment via YooMoney, Ozon Pay or Alfa-Bank.
Yes. Cloud models or your own server — your choice. Self-hosted means customer data stays inside your perimeter, and the open-source code can be audited.
Not on Start or Business — both run in our cloud. A dedicated server is only needed for the Team plan (on-premise with local AI) or for self-hosted install from GitHub: any Linux with Docker (from 2 GB RAM for cloud mode, from 12 GB + GPU for local AI).
Telegram, WhatsApp, web chat widget, telephony and the Android app. One assistant — every channel at once.
Yes, as part of the Team plan (on-premise). The assistant takes and makes calls via SIM7600 or SIP, recognizes speech and answers in voice — data never leaves your server.
Two paths — pick the one that fits.
Clone the code from GitHub and deploy on your own server. You need Docker and root SSH access.
We deploy on your server and tune it for your task. From payment to a working assistant — one business day.
Payment: YooMoney · Ozon Pay · Alfa-Bank
We'll reach out, answer your questions and activate your 14-day trial.