LLM agnostic voice-supported open-source chatbot solution for Telegram, built using NodeJS and MongoDB for the database. Bring your own API keys and deploy intelligent AI assistants with voice-to-voice capabilities.
Building a custom AI chatbot for Telegram often means:
Being locked into a single LLM provider (like OpenAI).
Facing soaring API costs as your user base grows.
Dealing with complex infrastructure just to get started.
Sacrificing privacy and data control.
Seamlessly switch between different LLM providers including OpenAI API specification and OpenRouter. Use the best model for each task without vendor lock-in.
Enable natural voice conversations with your AI assistant. Speak naturally and receive voice responses, making interactions more human-like and accessible.
Maintain conversation context as long as you need it. No artificial limits on conversation length or memory retention for truly intelligent interactions.
Built with extensibility in mind. Modify, enhance, and customize your chatbot however you like. Community-driven development with complete transparency.
Clone our GitHub repo and deploy to your favorite cloud provider (VPS, Railway, Heroku, etc.).
Plug in your preferred LLM API keys and customize your bot's personality and commands.
Add your new bot to Telegram channels, groups, or offer it as a one-on-one assistant.
Automate FAQs, moderate discussions, and engage members 24/7 in your Telegram groups and channels.
Provide instant, intelligent customer support and lead qualification directly through a familiar messaging app.
Create a personalized AI assistant, experiment with different LLMs, and build cool projects on a robust open-source foundation.
Free and open-source software with transparent pricing. Bring your own API keys.
"Crabo's agnostic approach let us cut our LLM costs by 60% by switching to a mix of models. A game-changer for our community."
"My eldest daughter used them in a school presentation, asking to distinguish between Crabo and a mediocre human."
Indie Hacker & Founder
I started dabbling with chatbots in early 2021, created a few small telegram bots. Then the OpenAI developer API was launched, someone wanted to have a chatbot on telegram as I talked to people on forums; That's how Crabo came into picture in August '22.
I've plunged into design, development, entrepreneurhip and writing, I always try to learn new things of interest and often find myself confused on what to learn next.
This was my first indie product, which also crossed six-figure in revenue and got a mention in a research paper. After an year of abandonment, I revived & planned to make it open source as it is a nice project for showcase.
It means Crabo is designed to work with any large language model. You can easily configure it to use OpenAI, Anthropic, open-source models like Llama 3, or others without changing the core bot code.
Yes, the software itself is 100% free and open-source (MIT License). You only incur costs based on the API pricing of the LLM provider you choose (e.g., OpenAI's GPT-4 costs, Anthropic's Claude costs).
Basic comfort with the command line and cloud services is recommended for self-hosting. However, our goal is to make the setup as simple as possible. When we launch Crabo Cloud, no technical skills will be required.
Absolutely. The MIT license explicitly allows for commercial use.
Join our growing community of developers and innovators.