1.Import your documents
Upload your PDFs, Word documents, or web pages. AgentMaurice's RAG system automatically chunks, vectorizes, and indexes them in pgvector.
RAG & Knowledge
Internal FAQ Assistant with RAG
Build an AI assistant that answers your team's questions using your internal documents — in minutes, without writing code.
agent.spec
No placeholder video. Product proof starts with the specification.
How it works
2.Define your Meta-Recipe
Create a declarative specification in Markdown + JSON that describes your assistant's behavior: system prompt, LLM model, RAG data sources, and response format.
3.Compile and test
AgentMaurice compiles your specification, validates dependencies, and runs tests automatically. Regression detection included.
4.Deploy in one click
Apply the Meta-Recipe to your deployment. The assistant is immediately available via chat, Slack, Telegram, or your custom interface.
Features included
Multi-document semantic search with entity extraction
6 LLM providers to choose from (OpenAI, Anthropic, Mistral, Gemini, OpenRouter, Ollama)
Accessible via web chat, Slack, Telegram, Discord, or API
Conversational memory for context continuity
Automatic content moderation via Mistral
Full observability via OpenTelemetry
The problem
Your teams waste hours searching for information scattered across dozens of internal documents. New hires ask the same questions, and the answers are buried in PDFs, wikis, or forgotten Slack threads.
The solution with AgentMaurice
In less than 30 minutes, you deploy a conversational AI assistant that relies on your own documents to provide accurate, sourced answers. No code to write — everything is declarative.
How it works under the hood
AgentMaurice's RAG pipeline automatically orchestrates:
Ingestion — Documents are split into intelligent chunks, enriched with entity extraction (people, organizations, concepts, technologies) and typed relationships, then vectorized and stored in pgvector.
Search — For each question, the MCP Brain server performs a hybrid BM25 + vector search to find the most relevant passages. The search mode is automatically selected based on the query type.
Response — The LLM generates a contextual answer based on the retrieved passages, with source citations. AgentMaurice memory ensures conversational continuity between questions.
Result
A fully operational FAQ assistant, deployed in production, accessible from all your communication channels — and it improves as you add more documents.
AgentMaurice
Request a Demo
Want to build your own?
Deploy your first AI assistant in under 30 minutes with AgentMaurice.