Demo

The platform

One tool for the entire lifecycle of your AI assistants

EasyRag brings together knowledge import, assistant configuration, publishing and administration. Here is how the pieces fit together, and what you do at each stage.

The 7-step journeySee the demo

Functional architecture

From document to sourced answer

Documents are analyzed, chunked and indexed once. For each question, the agent retrieves the relevant passages, queries your tools if needed, then writes an answer that cites its sources.

  1. Sources

    • Files (PDF, Office, text…)
    • Websites
    • Audio and video
    • Folders, FTP, SFTP

  2. Ingestion

    • Content validation
    • Extraction, OCR, transcription
    • Chunking and enrichment
    • Vectorization

  3. Knowledge

    • Corpora and priorities
    • Full-text index
    • Vector index (pgvector)
    • Corpus versions

  4. Agent

    • Hybrid search
    • Language model of choice
    • Tools: MCP, API, webhooks
    • Verified citations

  5. Publishing

    • Widget and public page
    • Conversation portal
    • Email, Slack
    • REST API
Administration: roles, SSO, organizations, licenses, GDPR, audit Monitoring: metrics, review, human handoff, reports
EasyRag functional architecture: from document to sourced answer.

Journey

Build and run an assistant, step by step

This journey reflects how the application actually works. Each step is handled by a specific role, so responsibilities stay separate.

  1. Connect a language model

    Register an API key from a provider (OpenAI, Mistral, Anthropic, Gemini) or the address of an internal model server. It is encrypted and can be shared.

    AdministratorDetails

  2. Build the knowledge base

    Create a knowledge base and feed it: files, websites, scheduled sync. Content validation can screen each document before indexing.

    EditorDetails

  3. Configure the assistant

    Describe the goal: AI suggests an identity, system prompt and welcome message. Choose the model, link knowledge bases with their priorities, add tools.

    EditorDetails

  4. Customize the behavior

    Tone, response language, guided flow, pre-chat form, escalation to a human, web search, handover to an agent.

    EditorDetails

  5. Test and compare

    Test window, side-by-side comparison of two assistants, benchmarking of several models on the same questions.

    TesterDetails

  6. Make it available

    Widget on your website, public page, internal chat portal, email, Slack or API.

    OwnerDetails

  7. Monitor and improve

    Metrics, conversation review, AI-powered analysis, escalations, reports. Every adjustment creates a new version of the assistant.

    SupervisorDetails

An integrated platform

Why a platform rather than a patchwork?

You can build a document assistant from separate components (vector database, RAG library, chat interface, monitoring tool). But then you have to connect them, secure them, maintain them and keep them evolving together.

EasyRag delivers these functions already connected, with shared administration: a newly added document can be queried right away, every conversation is monitored along with the sources it used, and a role applies everywhere.

  • Ingestion — Extraction, OCR, transcription, chunking, enrichment and embedding.
  • Search — Full-text and vector indexes, fusion, priorities, verified citations.
  • Agent — Model selection, MCP tools, APIs and webhooks, escalation.
  • Publishing — Widget, public page, portal, email, Slack, API.
  • Monitoring — Metrics, review, human handover, tickets, reports.
  • Administration — Roles, SSO, organizations, licenses, retention, audit, migration.
Which documents do I need to provide for a reimbursement request?

Attach the paid invoice and the signed request form 1. Above €500, prior approval from your manager is also required 2.

Sources

  1. 1Reimbursement procedure.pdfPDF“…every request must include the paid invoice and the signed form…”
  2. 2Internal memo no. 12.docxWord“…above €500, the manager’s prior approval is required…”
Interface mockup: every answer links back to the excerpts that support it (illustrative content).

Sourced answers

Verifiable by design

The assistant can only cite excerpts actually retrieved from your knowledge bases. If it finds nothing relevant, it says so.

Understand search and citations

Roles

Clear roles, enforced on the server

Configuring, testing, monitoring, answering users: six roles keep responsibilities separate. Permissions are checked by the server, not just hidden in the interface.

Security and governance
CapabilityAdminOwnerEditorTesterViewer
Manage users✓————
Configure assistants✓✓✓——
Manage corpora✓✓✓——
Keys, credentials, MCP✓✓———
View monitoring✓✓✓✓—
Test assistants✓✓✓✓✓
Permissions by role, based on the platform documentation (the “Support” role is reserved for conversation center agents).

Who it’s for

The right starting point for every profile

Business leaders

Assistants that put your documents to work and answer your teams or your end users, with no development project.

IT departments

Roles, SSO, retention, auditing, choice of models and installation in your infrastructure.

Consultants and integrators

White-label multi-organization, migration between instances, tools and APIs to build client solutions.

Developers

Documented REST API, JSON responses with sources, MCP tools and webhooks.

Self-hosting

Docker containers, self-hosted models for answer generation, data in your own database.

New users

Step-by-step guides and documentation organized by profile to help you get started.

Let’s see what EasyRag can do with your documents

Try the demo assistant, or tell us about your project: we’ll get back to you with a concrete feasibility assessment.

See the demoTell us about my project