Essential
€690
One source, one branded widget.
- One source: your site or your documents
- Branded chat widget, one line to install
- All four safeguards wired in
Answers from your documents.
Not from guesses.
Built on your site, your PDFs, your policies.
Demo shop assistant
Grounded on 6 documents · live
I answer only from this shop's six documents, and I name the source.
This one is real. Try it.
Every reply carries its source.
Ask it to ship to the US and watch it refuse instead of guess.
Answers only from your material.
Names the source. Every reply shows the document behind it.
Stops when unsure. It hands off to a person.
Honest fit
Where it is strong. And where it is the wrong tool.
Strong fit
Questions your documents already answer
Repetitive support and FAQ load
Internal knowledge spread across wikis and PDFs
Triage and routing of incoming requests
The wrong tool
Medical, legal or financial advice
Decisions your documents do not hold
Anything where a wrong answer is unrecoverable
A phone line. That is the voice agent
Your data
Your documents stay yours. That is the architecture.
Data handling
Art. 28 GDPR
Servers
EU only, in Falkenstein. No detour through US routing.
Training
Your documents never train a public model.
Control
Purge or re-sync the index at any time.
Contract
A data processing agreement (Art. 28 GDPR) is part of the contract.
Disclosure
It says it is an AI. Every answer is logged.
Read the data processing agreement and the sub-processor list.
The bench
Everyone else ships a prompt. Lucid AI Labs ships a memory.
A prompt-based assistant starts cold every single time. Someone pastes a block of context into a box and hopes it holds. What gets built here keeps a memory instead: every answer, correction and document becomes an entry that links to the entries around it, and the ones that get retrieved are weighted higher the next time. That network started life running the founder's own work, long before it ran anything sold. It is the reason these assistants stop guessing.
Left. Context goes in, an answer comes out, the run is thrown away. Ask it again tomorrow and it knows nothing about today.
Right. Each entry references its neighbours, so nothing is thrown away. The heavy line is the path retrieval keeps using, and it is the one reached for first.
This is the part still under active experiment. See what is on the bench.
Pricing
Priced in the open.
Signature
€1.290
Multi-source, in your voice.
- Multi-source RAG across all your material
- Tuned to your brand voice
- Lead capture plus human handoff
Bespoke
from€2.490
When it needs your other tools.
- Integrations: Shopify, Calendly, HubSpot, custom API
- Multi-language
- Everything in Signature
Then kept live
Care from €79/mo. Care+ at €149/mo adds deeper tuning.
60-day performance commitment: it answers correctly from your content, or your money back.
Questions
Common questions.
Q01What does RAG actually mean?
Retrieval-augmented generation. The assistant first retrieves the relevant passages from your own documents, then answers only from those. It declines when your material does not hold the answer.
Q02How is this different from a widget bought in an afternoon?
An off-the-shelf widget guesses to sound helpful. A grounded build answers only from your real sources and stops when it is unsure. Hand-engineered, with the four safeguards. No no-code glue.
Q03What happens when it does not know?
It stops. Below the confidence gate the assistant says it does not have a reliable answer and offers the next sensible step: a handoff, a callback, or a person.
Q04Where is the data hosted?
EU only, in Falkenstein. Your documents and their embeddings never leave the EU and never train a public model. A data processing agreement is part of the contract.
Q05Is there a guarantee?
Yes. A 60-day performance commitment, written into the contract: it answers correctly from your content, or your money back. The bar is at least 90 percent correct on a test set of your real questions, agreed in writing before you sign.