RagLeap Core
A free, self-hosted engine for building AI chatbots that answer questions using your own documents and business data, instead of renting a managed RAG service.
🔗 Visit RagLeap Core
Description
Chatbots that answer questions from a company's own documents — support tickets, product docs, internal wikis — usually run on something called RAG (retrieval-augmented generation): the AI looks up the relevant document before answering, instead of guessing from what it was trained on. Most services that offer this charge monthly per seat or per query and keep your documents on their servers. RagLeap Core is the same idea, but you run it yourself.
RagLeap Core is an open-source (MIT-licensed), self-hosted RAG engine extracted from a real production deployment that was answering actual customer questions, combining retrieval-augmented generation with multi-agent automation and voice capabilities. It ingests documents in 28+ formats, supports hybrid vector and full-text search, builds knowledge graphs via Neo4j, and offers 46 configurable "AI employee" roles. It connects to real-time business data from MySQL, PostgreSQL, MongoDB, and SaaS platforms like Salesforce, HubSpot, and Stripe, and ships channel adapters for WhatsApp, Telegram, Discord, and voice via Twilio, with support for 19 different LLM providers (OpenAI, Anthropic, Gemini, Mistral, Groq, DeepSeek, and others).
💬 Our review
The short version: RagLeap Core is a surprisingly complete, genuinely free alternative to piecing together your own RAG stack with LangChain — the tradeoff is you're the one running and maintaining it, and a few enterprise connectors are still missing.
LangChain is the obvious comparison as the dominant open-source RAG/agent framework, but it's a toolkit you assemble yourself, while RagLeap Core ships as a more complete, production-tested application with messaging-channel adapters and business-data connectors already built in — that's real time saved if your use case matches what it already supports. Managed options like Pinecone solve the vector-search piece well but leave you to build everything else (channels, agents, LLM routing) on top; RagLeap Core bundles all of that at zero license cost, which is hard to beat on price. The honest catch is entity extraction still leans on basic regex (expect some noise), a few common enterprise connectors (Snowflake, BigQuery, Airtable, Notion) aren't there yet, and being MIT-licensed means support is community-driven, not contractual — fine for a team with in-house engineering capacity, riskier if you need a vendor SLA.
📊 Global score
🤖 AI-enriched data
Open source (MIT), gratuit / auto-hébergé, sans clé de licence requise.
Pros
Licence MIT, entièrement gratuit, aucune clé de licence nécessaire
Code testé en production, extrait d'un déploiement réel répondant à des questions clients
Supporte 19 fournisseurs LLM (Gemini, Anthropic, OpenAI, Mistral, Groq, DeepSeek, etc.)
Adaptateurs multi-canaux intégrés : WhatsApp, Telegram, Discord, voix via Twilio
Cons
Extraction d'entités basée sur du regex basique, peut manquer des entités ou générer du bruit
Plusieurs connecteurs entreprise encore manquants (Snowflake, BigQuery, Airtable, Notion)