Comparison

IgnitionRAG vs AWS (Bedrock Knowledge Bases)

AWS offers Bedrock Knowledge Bases, OpenSearch Serverless and Kendra for building a RAG. IgnitionRAG provides the assistant ready to use, with cited sources, hosted in France on OVH.

Bedrock Knowledge Bases handles ingestion, indexing and retrieval; OpenSearch Serverless and Kendra are alternatives for the search layer. They are very good building blocks for an AWS team. IgnitionRAG also connects to AWS S3 and MinIO, and provides the complete assistant.

Cost: an estimate

A production deployment on AWS with OpenSearch Serverless (redundant collection, minimum 2 OCUs) comes to about €431 per month for 10 users and about €867 for 50 users, with Amazon Nova Lite and App Runner hosting in Paris. Kendra GenAI is an alternative at about €206 + €27 per month for search. Estimates based on official prices on 14/10/2026, excluding development and maintenance. Titan embeddings and authentication (Cognito): not priced. Dev-test or non-redundant modes are for prototypes.

What you have to build yourself

A production document assistant is more than vector search. You need to choose and configure the search layer, enforce per-document rights, display the citations returned by the API, develop the interface and authentication (Cognito), isolate each client and wire up CloudWatch. Each component has to be built, tested and maintained. IgnitionRAG ships them together: collections, answers that cite their sources, permissions, interface and monitoring.

When AWS is the better choice

If your data is already in S3, you have an AWS account, developers and a need to integrate into your applications, building on Bedrock is a solid choice.

Agencies and ESN

With the Agency/ESN licence, IgnitionRAG deploys on the client's Azure, AWS or GCP, or with IgnitionAI, BYOK, with white-label, on quote. White-label is not included in Solo or Team.

At a glance

Official pages aws.amazon.com (Bedrock, OpenSearch Service, Kendra) and the AWS Price List API (eu-west-3), checked on 14/10/2026. Converted at the ECB reference rate of 13/10/2026: €1 = $1.1188 (ecb.europa.eu). $482 → ≈ €431; $970 → ≈ €867; $230 → ≈ €206; $30 → ≈ €27. Titan embeddings and authentication (Cognito): not priced. Estimate under stated assumptions: low = 10 users, 2,000 documents, 5,000 questions/month; realistic = 50 users, 20,000 documents, 25,000 questions/month; ~4,000 input tokens and ~400 output tokens per question. Development and maintenance not included.

IgnitionRAGAWS (Bedrock Knowledge Bases) (to build)
Getting startedReady-made platform: collections, chat, citations, connectorsComponents to assemble and maintain
PriceSolo €59/month excl. VAT · Team €249/month excl. VAT, AI credits includedProduction estimate: ≈ €431/month (10 users) to ≈ €867/month (50 users), excluding development
Model API keyNone on SaaS (Solo, Team): AI credits included, no key to provide; BYOK with the Agency/ESN licenceYour keys and your model billing
Connected sourcesWeb (scraping), Azure (SharePoint, Tables, Blob), AWS S3 + MinIO, GitHub, Hugging Face; WhatsApp (Unipile, beta)S3, SharePoint, Confluence… on the Knowledge Base side
Source citationsIncludedTo implement and display
Permissions and multi-organisationBuilt inMetadata filters / IAM and per-client KB to design
Chat interface and authenticationIncludedTo develop; Cognito
HostingFrance, OVH (SaaS); Agency/ESN licence: client's Azure, AWS or GCP, or with IgnitionAIYour AWS account, region of your choice (e.g. Paris for OpenSearch)
Agencies / ESNAgency/ESN licence: deployment on the client's Azure, AWS or GCP, or with IgnitionAI, BYOK, white-label, on quote. Not in Solo or Team.To build per client
Trial14-day free trialLimited free tiers (e.g. Kendra 750 h over 30 days)

FAQ

Is it cheaper to build it yourself?

In production, no: under our assumptions, an AWS deployment (OpenSearch Serverless + Bedrock) comes to about €431 per month for a small team and €867 for 50 users, before development and maintenance. IgnitionRAG: Solo €59/month excl. VAT, Team €249/month excl. VAT, AI credits included.

When is it better to build it yourself?

If your data is in S3 and your team already develops on AWS, with time to build and maintain the application.

Do I need a model API key with IgnitionRAG?

No for Solo and Team: AI credits are included. The Agency/ESN licence works with BYOK, with white-label, deployed on the client's Azure, AWS or GCP or with IgnitionAI, on quote. White-label is not included in Solo or Team.

Does IgnitionRAG work with WhatsApp?

Yes, via Unipile, in beta.

How can I try it?

Join early access: 14-day free trial.

Deploy AI agents that work for your company.

Connect authorized knowledge and tools, then put an agent to work where your team needs it.