Less searching.
Get more done.

Find the answer in a contract. Prepare for the next customer conversation. Have a recurring summary ready for you. Kivanto brings your knowledge and your work together, with AI that can access your documents.

Your knowledge. Your workflows. On your infrastructure.

Download for free ↓Book a consultation

Kivanto Solo · Free for your computer

> kivanto / a workspace for your knowledge

[01] Find answerswith evidence

[02] Support customerswith context

[03] See connectionswith sources

[04] Prepare routineson a schedule

[05] Create contentwith suitable models

> foundation: your company knowledge_

Sound familiar?

01The information you need is somewhere between a shared drive, a document store and your CRM.
02Before a customer call, you are missing the last commitment, the right document or the next step.
03Recurring summaries, notes and follow-up tasks start from scratch every time.
04Your AI knows the question but lacks the documents and the context.

Aus vorhandenem Wissen wird your next step.

What did we promise in the proposal? Who is part of the project? And which source supports the answer?

Documents, contacts and connections in one platform.

Kivanto in everyday work

What would you like to get done today?

Start with a question or a task. Your documents, customer information and connected tools provide the context.

Move forward with one question.

“What did we agree — and where does it say so?”

Ask your documents instead of searching folders one by one. Kivanto combines search and chat: get an answer with references and read the original source. Prepare for a conversation or get up to speed on a topic.

Start with a document, a scan or a recording

Drop a PDF, Word document, spreadsheet or presentation into the chat and ask about it. Images and scanned PDFs use a configured vision model; audio recordings use a transcription model. Choose whether an attachment stays in the conversation or becomes reusable project knowledge.

For important statements, turn on “Thorough” to request additional checks against original sources. See supporting evidence, contradictions and unresolved points. The agent can use a local calculator for arithmetic. You remain responsible for assessing the answer.

When your own documents are not enough, enabled web research can add external sources. It must be configured for your deployment and sends search queries to external services. Reopen conversations later or share them with authorised colleagues to read or continue.

Understand what a change could affect.

“If this supplier fails, what depends on it?”

See how people, companies, projects and documents connect. Impact analysis follows recorded dependencies and shows potentially affected items with an evidence trail. Know where to take a closer look first.

Follow a connection back to the evidence

In the knowledge graph, move from a subject to its relationships and then to the relevant source passage. Custom concepts and relationship types help represent your industry. Reviewed additions and structured exports support further use.

The analysis shows possible dependencies in visible, indexed knowledge. It does not predict actual damage. Check whether stock, alternatives or other agreements change the outcome in your specific situation.

Be ready for the next customer conversation.

“What is still open, what did we promise and who owns the next step?”

Open the customer record to bring contacts, sales opportunities, tasks, notes and related documents together. Keep the next step visible alongside the deal. Personal views and sales summaries help you prioritise today’s work.

Fit customer work to your processes and existing systems

Organise opportunities on a board or in a table. Custom sales stages, fields, filters, lists and dynamic contact groups reflect how you work. Attachments and change history stay with the customer record. Use chat to look up CRM data and prepare changes for your approval.

A configured Microsoft 365 connection brings email subjects, previews and calendar events into your activity history. Brevo campaign reports show available delivery and engagement figures. Import Twenty data with field mapping, review conflicts and preview supported changes before writing them back. Connections and permissions are configured for your actual accounts.

CRM workflows can respond to a change by creating a follow-up task or note, updating a sales stage or preparing an email draft. Conditions and delays control the sequence; a test run previews the steps. Open email drafts in your email application to send them.

Set up recurring questions once.

“Summarise this project’s status every Monday.”

Save an instruction as a routine. Kivanto runs it at the time you choose and saves the result as a conversation. Open the summary, check its sources and continue working from there.

Use schedules and events that fit your day

Run once, hourly, daily, on weekdays or weekly, in your time zone. Prepare a routine for approval from chat, run it manually or pause it. History shows results and failed runs.

Events from connected systems can trigger a routine through a webhook. The sender must be able to reach the Kivanto instance. For every routine, Kivanto must be running, but the browser can be closed. Required approvals for proposed actions still apply.

Turn an idea into material you can use.

“An image for the post, a voice for the text.”

Use Media Studio to create images, spoken text and audio with configured models. Turn an uploaded recording into a transcript. Preview, listen to and download your results.

Choose models for the result you need

Use suitable enabled models through services such as OpenAI-compatible providers, fal.ai or Higgsfield. Available image, audio and speech functions depend on the configured model. Find your jobs and completed results in your personal history.

Text or files sent to external providers are processed there and may incur charges. Media Studio does not currently generate video; transcription works with uploaded files.

Bring your knowledge into the work you do with AI.

“Use our documents and prepare the next step.”

Connect your AI assistants to shared project knowledge. They can look up information and, with the appropriate permissions, edit files or CRM data. In Kivanto chat, review proposals before approving changes.

From research to drafts and reviewed code changes

Claude Desktop, Claude Code and Codex can use Kivanto through MCP, an interface for AI tools. Each client receives its own revocable credentials. The Kivanto agent can also call read-only tools from configured MCP connections and bring their results together for your question.

Prepare a note, a structured analysis or a new text file as a draft. For development work, the project agent can search code, inspect Git changes and propose specific file edits. With a configured runner, approved tests and builds run in an isolated project copy.

Custom interfaces and extensions can support your business processes. Access to data and actions is defined during setup; connecting an AI client does not give it unrestricted access to every project.

Preview of upcoming versions

From a request to a result.

These new capabilities are implemented in the current development version. They are not yet part of the verified stable download. We will agree the appropriate version and setup for your use case.

From a conversation to a usable document.

“Turn this into a report and update the budget spreadsheet.”

Prepare Word documents and Excel spreadsheets, or make targeted changes, directly in chat. Review the content and proposed changes, approve the file and continue working with it. There is no need to copy individual answers into a document.

What you can review before approval

Word supports text, headings and tables; Excel supports edits to cells, formulas and worksheets. Content previews show the draft and proposed changes. HTML drafts can be viewed as a rendered page.

File editing runs locally and does not require an Office installation. Your selected AI model may still be an external service. Previews do not reproduce the printed Office layout; new Excel formulas are calculated when opened in Excel.

Hand over a task and stay in the conversation.

“Make this change in the project and show me your progress.”

Start OpenCode tasks directly in Kivanto chat. The agent receives your request, conversation context and suitable attachments. Follow its steps, answer questions and respond to approval requests without switching windows to coordinate the work.

Connect your own assistant to project knowledge

Direct execution requires Kivanto Solo, a local project folder and a configured OpenCode installation. OpenCode uses its own models, tools and permissions; changes are made directly in the project folder. Stopping a task keeps changes already made. External agent tasks cannot yet be scheduled as routines.

If you prefer working in your existing assistant, guided setup connects OpenCode and Pi Agent to Kivanto. Cline has a manual connection option; Claude and Codex remain supported. Access to project knowledge is separate from running OpenCode directly in chat. Each connection can be revoked individually.

Use your own examples to prepare recurring decisions.

“Which cases should we look at first?”

Use examples reviewed by your team to distinguish document types and surface signals about sales opportunities, customer risks or possible project delays. Local specialist models help with the initial selection. Your team reviews the signals and decides what to do next.

Learn from your own data and check the results

In “Learning & models”, collect confirmed examples, train models and compare their results. Activation requires separate training and evaluation data and passed quality checks. Without suitable examples, there is no trained model for your task.

These specialist models are trained and run locally. They can also support document extraction, relationships and narrowly defined source checks. Suitable simple calculations and knowledge queries work without a generative model call; other questions continue to use the configured language model.

Signals are not guaranteed predictions or actual probabilities of closing a deal. Suggested relationships are not automatically accepted as facts. Confirmed training examples are stored separately and need to be included in any later data deletion.

Choose a model for the task and your budget.

“Which model suits this work, and what does it cost to use?”

Compare local models and configured providers in one place. Where providers supply the data, model information and token prices are visible during selection. This helps you weigh suitability against expected costs.

Understand provider choices and cost estimates

Setup templates support European and international providers as well as local model servers. Context limits and prices can be retrieved from supported providers. For other services, enter the model and prices yourself. Required capabilities are configured for each model.

A provider’s headquarters do not establish where your request is processed; the selected service and your agreement determine that. Price information and the usage dashboard help you estimate costs but do not replace billing. Missing prices do not mean free usage.

Connect knowledge. Understand relationships.

Start with your documents. Keep working together.

Start on your own with Kivanto Solo or set up a shared instance for your team. You decide which knowledge to include and who can work with it.

Keep using the knowledge you already have.

Start with a folder, your project files or a connected document store. Upload multiple files, organise them in the file browser and open search results directly in the preview.

Which sources can I use?

Configure sources from local folders, mounted network drives, Paperless archives, Calibre libraries and ZIM reference collections. A local source folder is imported read-only; synchronise it to pick up later changes. MCP provider profiles include HubSpot, Pipedrive, Monday.com, Brevo and Composio. Accounts, permissions and available read tools are checked for each connection.

Share knowledge with the right people.

Organise collaboration by project. Give a colleague access to the documents they need or share a conversation to read and continue. Personal spaces stay separate.

How do teams get started?

The Server edition supports shared use with roles and read/write permissions. Existing Microsoft or LDAP sign-ins can be connected through a configured identity service. Solo remains the free edition for one person on their own computer. Both interfaces are available in German and English.

Use an AI model that fits the task.

Choose an enabled model that fits your task. Run models locally or configure an external provider. Select suitable models specifically for images, audio and text recognition.

What does this mean for data and costs?

With external models, the content needed for a request leaves your instance. Your personal usage dashboard shows available chat token counts and cost estimates based on configured model prices, down to individual calls. Media generation and other external services are not fully included. Missing measurements are shown as unavailable.

Start with a manageable task.

Install Kivanto Solo for Windows, Mac or Linux and follow the setup wizard. Choose an AI provider and add your first documents. For team use, we help with setup, data connections and onboarding.

How do you keep control of your setup?

Desktop packages include the required runtime and a local database. Use the status icon to start or stop Kivanto and reopen model settings. For teams, the shared instance runs on your infrastructure; data transfer, backups and operational procedures are agreed during setup.

Discuss your workflow

Make everyday responsibilities traceable.

When you need to review access or handle a privacy request, you need traceable records. Kivanto helps you review events and prepare decisions.

Trace what happened when questions arise.

Review who requested project access, when it happened and whether it was allowed. Recorded response, research and settings events add context. Filters and CSV/JSON exports help you assemble relevant entries for a review.

Prepare privacy requests in an organised way.

Record responsibility, deadlines, identity checks and relevant data references in a case. Document your assessment and export the review package. Actual data discovery, disclosure or deletion is carried out separately.

Grant access deliberately.

Projects, roles and sharing settings control access to your knowledge. Personal spaces remain separate. Connected AI assistants also use assigned permissions and revocable keys.

Choose how processing is configured.

Run Kivanto on your infrastructure and choose your data sources and AI models. Local models can keep AI processing within your own environment. External providers are connected deliberately.

GDPR in practice.

For your deployment, you define processing purposes, legal bases, retention and deletion periods, and the required agreements with service providers. Kivanto provides technical access-control and traceability features; GDPR compliance depends on your configuration and organisational processes.

Read the GDPR ↗

Four steps to Kivanto

Define your use case

Initial consultation

We discuss the knowledge your team needs, where it is stored today and the tasks Kivanto should help with.

Choose your setup

Infrastructure & models

Together we choose the installation environment, AI models and team access. We then prepare your proposal.

Connect your knowledge

Sources & projects

We configure the agreed data sources, projects and permissions, then check the results using your documents.

Get your team started

Training & handover

Your team learns to use search, chat and the knowledge graph. If needed, we connect your AI assistants and provide ongoing support.

Daniel Schlager, software developer and specialist in digitalisation and AI

Daniel Schlager · Developer of Kivanto

Daniel Schlager.
The developer behind Kivanto.

Software developer and specialist in digitalisation and AI.

Daniel Schlager has many years of experience in software development. He builds applications, connects existing systems and helps businesses bring digital processes and artificial intelligence into their daily work.

From software architecture to everyday business use.

With Kivanto, he brings this experience together in his own platform for company knowledge and AI. As its developer, he is your direct contact for implementation, data integration and use within your team.

Development and rollout from the same person.

Daniel Schlager develops Kivanto in Salzburg and supports you from the first use case through setup to team adoption. He combines technical experience with an understanding of your business processes.

More about Daniel Schlager ↗

Frequently asked questions

01How does Kivanto help with everyday work?+

Kivanto connects your knowledge to the tasks you want to accomplish: ask about documents, check claims against sources, explore dependencies, prepare customer conversations, organise follow-up tasks and schedule recurring summaries. Media Studio and AI assistants support further work. The instance runs on your infrastructure, with models and connections configured for your needs.

02How does Kivanto Solo differ from team use?+

Kivanto Solo is free for one person on their own computer, including professional use. A shared instance for multiple people requires the Kivanto Server edition under a separate agreement. Setup, data integration and support are offered individually.

03Where does Kivanto run, and where is our data processed?+

Kivanto runs on your own infrastructure, either locally on a computer or on a company server. You can configure local models or an external model provider for AI processing. With an external provider, the content needed for each request is sent to that provider. Models and data sources are selected during setup.

04Do we need to replace our files and existing systems?+

Start with existing folders, network drives and supported document stores. Local source folders are imported read-only and synchronised on request. MCP connections add read access to other services. Dedicated CRM connections for Twenty, Microsoft 365 and Brevo each have their own scope. Credentials, permissions and data mappings are configured and verified for your accounts.

05Does the AI change or send anything without my approval?+

In Kivanto chat, file and CRM changes are prepared as proposals for human approval. The same applies to code edits and configured test commands. Explicitly activated CRM workflows execute their configured actions automatically. They create email drafts but do not send messages. External AI clients operate with the permissions you grant and their own approval rules.

06Can I keep using my existing AI assistants?+

Yes. Claude Desktop, Claude Code and Codex, for example, can access shared knowledge in Kivanto through MCP. Each client receives its own revocable credentials with defined permissions. The Kivanto agent can also use read-only tools from configured MCP services. Choosing a model in Kivanto does not automatically determine which model an external assistant uses.

07Can Kivanto edit Word and Excel files for me?+

In the current development version, the built-in agent can create DOCX and XLSX files and make targeted changes. You review a content preview and proposed changes before approving the file. Editing does not require an Office installation. New Excel formulas are calculated when opened in Excel. This capability is presented as a preview and has not yet been verified in the stable download.

08Does Kivanto automatically learn from all my data?+

The new local specialist models are deliberately trained on confirmed examples. You review the results and activate a model only after the required quality checks. There is no automatic training on all your documents. Models and examples remain tied to the respective user and project context. This learning area is part of the current development version; reliable use requires suitable data of your own.

09Who can see which content?+

Kivanto organises knowledge into projects with members and roles. Personal spaces are separate from shared content. Access by AI assistants is also tied to user permissions. Data sources and sharing settings are configured to suit your teams.

10Do routines run when the browser is closed?+

Yes, as long as the Kivanto instance is running. A routine starts at the scheduled time and saves its result as a conversation. It does not wake a sleeping or switched-off computer. Webhook triggers need an instance the sender can reach. Actions that require approval are not automatically approved by routines.

11What records are available for internal reviews?+

The audit log records allowed and denied project access, along with other captured response, research and settings events. Authorised users can filter and export entries. Permission to access does not prove that an action was completed; the log is not a complete, tamper-proof archive. Privacy cases support manual preparation for a decision but do not automatically disclose or erase data.

12How does Kivanto support GDPR implementation?+

Self-hosting, role and project permissions, and traceable access decisions support your technical and organisational privacy measures. GDPR compliance also depends on processing purposes, legal bases, configuration, provider agreements, and retention and deletion policies. Automated access and deletion workflows across all data stores are not yet fully available. Audit logs themselves can contain personal data and must be included in your privacy policies and procedures.

13How do we get started, and what does Kivanto cost?+

You can download Kivanto Solo for free and use it as an individual on your own computer. Central company setup, data integration and support are offered through an individual proposal. Infrastructure and external AI models may incur additional costs.

Let's put your knowledge to work.

Start for free with Kivanto Solo on your computer. For team setup and connecting your data sources, we provide personal support.

Download for free ↓office@danielschlager.ai

Less searching.Get more done.