Sign in

Products

We have built six tools so far, which we run ourselves on servers in the EU. All of them can be used in komplai, and redact, lens and secret can also be used free of charge without an account.

chat

An AI chat that can use all of komplai’s tools

  • It can look things up in registers such as CVR and Retsinformation, and search the organisation’s own documents with Index.
  • With Redact switched on, the personal data Redact finds is replaced with labels such as [PERSON-1] before the model sees the text. We put the names back in the answer.
  • For recurring tasks, people can set up assistants and skills, each with its own instructions.
  • Each answer shows which model wrote it and where it ran.
Runs
On our local servers in Denmark, in our EU cloud, on the organisation’s own servers or with an AI provider the organisation has chosen.
Available
In komplai and through the API.
Example

Write a short reply to Maja Eksempel saying we have received her request for access to documents, and refer to the deadline in the Access to Public Administration Files Act.

The model sees

Write a short reply to [PERSON-1] saying we have received her request for access to documents, and refer to the deadline in the Access to Public Administration Files Act.

ToolsRedactRegister lookup · Retsinformation

Dear Maja Eksempel, thank you for your request for access to documents. We have received it and will reply as soon as we can. If we cannot reply within 7 working days, we will tell you why, as section 36(2) of the Act requires.

Qwen3.8 27B · Local servers · Denmark · Redact on

redact

Removes personal data from text and PDFs

  • In text, names and numbers are replaced with labels such as [PERSON-1]. The text can then go to an AI service, and the names can be put back afterwards.
  • The redacted PDF is saved as images, page by page. There is no text under the black boxes for anyone to copy or extract.
  • Regex finds numbers such as CPR numbers, and NER finds names and addresses. Each run states which methods were used and where they ran.
Runs
Regex runs in the browser and again with NER on our servers in Germany. In komplai, NER can also run on our local servers in Denmark, and a language model can look for personal data on our local servers or in our EU cloud.
Available
Free on redact.komplai.dk. In komplai, with the API and Word as an add-on.
Example

Note on Maja Eksempel (found with NER), CPR 010180-1234 (found with Regex). She is the only night nurse on the Example ward (found by the agent from the context) and is on sick leave after knee surgery (found by the agent from the context).

The colour shows which method found each item.

Note on [PERSON-1], CPR [CPR-1]. She is [KVASI-1] and is on sick leave after [HELBRED-1].

  • redact.komplai.dk

    Free for text and PDFs, no account needed.

  • Redact in komplai

    For text and PDFs, with a choice of model and of where it runs.

  • Redact API and Word

    Redact can be called from the organisation’s own systems through the API and used directly in Word.

Redact AgenticAlpha

Redact Agentic is an AI agent that reads the whole text and decides for itself which tools to use. It can also find details no pattern can match, such as a person being the only night nurse on a ward or having a particular diagnosis.

Redact Agentic is in Alpha and can run on local infrastructure in Denmark or in our EU cloud.

The agent

Reads the text and picks its tools as it goes.

  • NER modelsSeveral models that find names, organisations and addresses.
  • RegexFixed patterns such as CPR and phone numbers.
  • Its own regular expressionsThe agent writes new patterns when the text calls for it.
  • LookupsChecks registers such as CVR to see whether a finding is correct.

Finally, the agent reads the finished text again to see whether it missed anything.

lens

Shows what changed between two versions of a contract

  • Lens compares two Word files, two PDFs or one of each. A moved section or a new page number in a PDF does not count as a change.
  • The comparison happens in the browser without AI, and the files are not sent to us. Scanned pages are also read with OCR in the browser.
  • It can also answer questions about the contract or summarise it. The answer points to the places in the document it is based on.
Runs
Comparison and OCR run in the browser. Questions and summaries run on our local servers in Denmark. On lens.komplai.dk, people can also use their own API key with another AI provider.
Available
Free on lens.komplai.dk, no account needed. Also in komplai.
Example
Version 1 · PDFVersion 2 · PDF

§ 3 Price. Eksempel ApS pays DKK 120,000.

§ 3 Price. Eksempel ApS pays DKK 135,000.

Changed

§ 7 Confidentiality

§ 9 Confidentiality

Moved, not changed

Page 3 of 12

Page 4 of 14

Page numbers do not count

1 change

index

Review of large matters and entire mailboxes

  • Index brings in email and documents from entire mailboxes and archives and can work with millions of documents in one matter.
  • Semantic search finds documents by their meaning, and AI groups them into clusters by topic.
  • In a review, every document gets the same questions, and the answers sit in a table, each pointing to the place it is based on.
  • People can ask across the whole matter or let an agent carry out tasks in several steps, such as finding every agreement with a particular clause. The agent shows its plan and starts only once it is approved.
Runs
The documents are stored on our servers in Germany. When AI is switched on, it runs on our local servers in Denmark.
Available
As an add-on in komplai, with no limit on the number of users.
Example
Review: two questions for every contract
ContractNotice periodLiability cap
IT operations framework agreement6 months§ 14.2One year’s fees§ 18.1
Subcontractor agreement3 months§ 9No cap§ 11Differs from standard
Addendum 2Not foundAs the main agreement§ 2

From mailbox to disclosure

  1. CollectionPST, MBOX, EML and MSG, Teams messages, Microsoft 365 exports, zip files and scans. Attachments are unpacked, and scanned pages are read with OCR.
  2. ProcessingDuplicates are found, emails are grouped into threads, and every document gets its custodian, sender, recipients and date.
  3. AnalysisSemantic search, topics in clusters, who wrote to whom and when, and timelines of the matter.
    Topics in a matter
    • Termination214 documents
    • Payment87 documents
    • Defects41 documents
  4. DisclosureCourt extracts and exhibit lists, document lists with deadlines for access requests, chronologies, search term reports and export to Excel and Word. Personal data can be blacked out with Redact first.

run

Share apps built with Claude Code, Codex and Aura with colleagues

  • Many people have started building small apps with Claude Code, Codex or Aura but have nowhere to put them so colleagues can use them. With Run the assistant is asked to publish the app, and it gets its own address on komplai.run.
  • The app can be shared with the organisation or publicly. Its code runs in the user’s browser and never on our servers, and an app never gets more access than the person using it.
  • An administrator approves what each app may do, such as store data, use AI or write to the event log. When needed we also set up network monitoring, a firewall and other guardrails.
Runs
App files are stored on our servers in Germany. Data the apps save is encrypted in our database in the EU.
Available
Included in komplai, and as run/dedicated on the organisation’s own server.
Example

Claude Code

Publish the meeting room booking so colleagues can use it.

Published at moedelokaler.komplai.run. Only people in the organisation can open it.

moedelokaler.komplai.runOrganisation only

What the app may do

  • Show who made a bookingApproved
  • Save bookingsApproved
  • Write to the event logApproved
  • Use AINot requested
  • Run in komplai

    5 GB for apps on our shared servers in the EU, typically enough for several hundred small apps. Apps can also be published straight from komplai.

  • run/dedicated

    A server in the EU that runs only the organisation’s own apps, so they share capacity with no one else.

secret

One-time links for passwords and anything else that should not be left in an inbox

  • The text is encrypted in the browser. The key is after the # in the link and is never sent to us, so we cannot read the text.
  • As soon as the recipient clicks to see the text, we delete it from our server. It can be seen only once.
  • Email scanners that open links before the recipient does cannot use up the link. The text is fetched only when a person clicks.
Runs
The text is stored encrypted on our servers in Germany until the recipient opens it or the link expires.
Available
Free on secret.komplai.dk, no account needed. Also in komplai.
Example

secret.komplai.dk/s#k3Jx9wQe.Zp7rT2mLc8VnYq4F

  • id: k3Jx9wQe
  • key: Zp7rT2mLc8VnYq4F, never sent to the server
  • Can be opened once
  • Expires after 24 hours
  • Notification when the link is opened

Price and setup

We set up komplai with each organisation and agree on operations and pricing with each one. Write a few lines about the task and we will get back.

If you'd rather skip the form, write straight to hej@komplai.dk

See prices

We only use what you write to reply to you. How we handle the information