amaiko vs Onyx: Self-Hosted Control or Sovereignty Without the Infrastructure? (2026)

What Onyx does genuinely well

Onyx — formerly Danswer, so if you arrived here looking for a Danswer comparison, you are on the right page — has earned real credit, and it gets that credit before a single word of criticism.

Genuinely open source. Onyx is MIT-licensed. You can read every line, audit the security yourself, fork the project if its direction stops suiting you, and never be hostage to a vendor’s roadmap. In a market saturated with “trust us” compliance claims, source code you can inspect is a different category of trust.

Self-hosting means real sovereignty. Onyx runs wherever you put it — including a server in your own German datacenter. Your data doesn’t leave the building: not contractually protected, physically present. No jurisdiction debate, no adequacy decisions, no fine print. Among all the “GDPR-friendly” marketing in this space, this is the genuine article.

Breadth and model freedom. More than 40 connectors — Microsoft Teams included — index your scattered tool landscape, and the platform is model-agnostic: you choose which LLM does the answering, including one your data-protection officer can actually live with.

And the price is honest: $20 per user per month for the managed cloud, or free if you self-host. For a capable IT team that wants full control over its AI stack, Onyx is genuinely attractive.

So why does this page exist? Because “a capable IT team” is doing an enormous amount of work in that sentence — and because enterprise search and an AI assistant are not the same product.

Five structural differences

The amaiko vs Onyx question is not open source vs proprietary. It is two questions: who operates the infrastructure, and what does the software actually do once it runs?

Self-hosting is a job, not a feature

“Free if you self-host” is true the way a free puppy is free. Someone has to deploy Onyx, secure it, patch it, monitor it and scale it — not once, but continuously, for as long as the company depends on it. That is an ongoing engineering commitment, and most Mittelstand IT teams are staffed to keep the ERP alive and the printers working, not to operate an AI platform on the side. The engineer who runs your Onyx instance is an engineer not doing whatever you actually hired them for. amaiko’s answer is that there is nothing to operate: it is a managed service hosted 100% in Germany, and your IT team’s involvement ends roughly when the Teams installation finishes.

Enterprise search is not an assistant

Onyx is enterprise search at its core: you ask, it retrieves. Genuinely useful — but every insight still sits behind a question someone has to think to ask, and the questions nobody thinks to ask never get answered. amaiko works the other way around. It monitors your company’s signals and acts first: your morning briefing is ready before you open the laptop, your inbox is triaged before you read it, meeting follow-ups appear without anyone asking. A search box saves you minutes when you use it; a proactive assistant saves you the minutes you didn’t know you were losing.

Retrieval is not memory

Onyx indexes documents and retrieves them at query time. It does not accumulate knowledge — there is no persistent, self-learning memory that gets smarter about your company over time. amaiko builds exactly that: a persistent corporate memory of decisions, context, who knows what, and why things were done the way they were. When an employee leaves, their context stays. The measurable effect for amaiko teams: 35% less time spent searching and onboarding up to 57% faster, because new hires inherit a memory instead of an empty search box.

A Teams connector is not a Teams-native assistant

Onyx’s Teams connector indexes your Teams content as one more data source — a real capability, properly credited. But the product itself lives in its own interface: another tab, another login, another behavior change to manage. amaiko is Teams-native. The assistant lives where your team already works, briefings arrive in the chat window that is already open, and a network of 24 specialized AI agents coordinates behind it — for meetings, email, research, knowledge linkage and more. See how the agent network operates.

Setup all the way down

Even with the infrastructure question solved — say you pay for Onyx Cloud — the work isn’t done, because end users still need setup: sources to connect, permissions to configure, a new tool to learn. That is an adoption project with an owner and a timeline. amaiko’s onboarding is a single Teams chat. You install it, say hello, and it starts working — no training program, no rollout plan, no change-management deck.

Same sovereignty, zero infrastructure

The reason Onyx appears on a German company’s shortlist is almost always data sovereignty — and credit where due: self-hosted Onyx genuinely delivers it. But it delivers it at the price of becoming an infrastructure operator, and the cloud version at $20 hands the data question back to a US vendor’s cloud, which rather defeats the reason you shortlisted it.

amaiko’s position is simple: the same compliance outcome, zero infrastructure. 100% German hosting and ISO 42001 certification — the management standard for AI systems — with German-speaking support when something needs clarifying. You get the outcome the self-hosters are after, without staffing the project. For a Mittelstand company, that is the difference between a compliance decision and a hiring decision.

The pricing reality

Onyx Cloud costs $20 per user per month — a fair price, and cheaper than most US enterprise platforms. Self-hosted Onyx is free at the license level; the real bill arrives as engineer-hours: deployment, hardening, updates, scaling, and the standing readiness to fix it when it breaks. For most companies under a thousand employees, the “free” option is the expensive one.

amaiko starts at €19.92 per user per month, billed annually — managed, German-hosted, nothing to operate. A comparable sticker price to Onyx Cloud, with a radically different operating model underneath.

Who should choose which

Honest segmentation, no sales reflex.

Choose Onyx if you have a platform team with genuine DevOps capacity, MIT-licensed code you can audit and fork matters to you, full control over models and infrastructure outweighs convenience, and your core need is search across a wide tool landscape. In that scenario, Onyx is a strong, honest choice — and the self-hosting burden is a price you have consciously decided to pay.

Choose amaiko if you want the same sovereignty outcome — your data in Germany, under European law — without deploying, securing, updating or scaling anything, and you want AI that acts before you ask, permanently remembers what your company learns, and lives natively in Teams instead of a search box in another tab.

Run both if your IT team already operates Onyx happily as internal search and you want proactive, memory-backed assistance in Teams on top. They are not fighting over the same seat.

If you are surveying the wider field, our roundup of Onyx alternatives covers the other contenders. And if you would rather see the zero-infrastructure version of data sovereignty than read about it: book a demo — it takes one Teams chat to show you.