Choose your plan.

Yes. Private individuals get knowmind without payment details and without a trial ending. The Private plan includes up to 2,500 memories — plus the full API, the CLI and the MCP server, so you can use knowmind directly in Claude Desktop, Cursor, ChatGPT or a script of your own. If you use the tool seriously and would like to donate, you can do so in the dashboard or through the footer.

The Team plan includes five people in the base price, the Business plan ten. Every additional person in the same workspace costs 15 euros per month on top. Everyone shares the same body of knowledge; the role is configurable per person (owner, admin, member, reader).

You store your own provider key — for Claude, ChatGPT, Gemini or Mistral, for instance. knowmind uses it only for your requests. You pay the AI costs directly to the provider, not to us. We add no margin on inference. In industry jargon this model is called BYOK (bring your own key).

Yes. API, CLI and MCP server are included in every plan — including the free Private plan. After `knowmind login` you can immediately create and query memories over the REST API, add the MCP server to Claude Desktop or Cursor, or write your own scripts against the interface. The only limits on Private are the 2,500-memory cap and a rate limit of 30 requests per minute.

Team (99 euros, five seats) is the entry point for teams and small companies wanting a shared body of knowledge, roles and invitations. Business (349 euros, ten seats) pays off from around ten people, or when you need a longer audit log (24 instead of 12 months), a single sign-on connection and prioritised support. Both share the same functional core; every additional person costs 15 euros per month in either plan.

On a German server in Nuremberg at Hetzner. If you store your own key and pick a US model, the AI requests go directly to that provider — that is your decision. An EU-aligned variant is to connect Mistral, Aleph Alpha or a local model through LM Studio or Ollama.

Yes. The dashboard has an export button that downloads all memories and documents as a JSON file. With it you can take your content to another provider or process it locally. You request full account deletion by email to info@schuebeler-consulting.de and receive confirmation within 30 days.

Committed response times, agreed availability and a named contact are part of the Enterprise plan. For companies with concrete requirements on response time, availability and a named contact, Enterprise comes with a data processing agreement and, on request, an on-premise variant.

knowmind stores your memories and returns them on request. The AI's actual answer runs through your own provider key, directly at that provider. We do not check the content of AI answers and cannot be liable for their accuracy. For safety-critical decisions, always check AI answers against the original sources.

Right after installation. As soon as the CLI has a valid token from `knowmind login` and the MCP server is registered in your AI tool, all tools are available in the `mcp__knowmind__` namespace. You do not have to create a memory or click through a configuration. The one thing worth doing is running `knowmind sync <folder>` once on an existing notes folder, so the knowledge store has content and recall returns useful hits from the first question on.

That is the background sync of your files. On Windows it briefly shows a console window that closes by itself. Your data is not affected.

No. After signing in once, a session stays valid for 30 days (browser cookie). Only when you actively sign out or clear cookies do you need a new magic link. If you find that annoying, you can also set a password (available in your profile) or work through the CLI with an API token that stays valid as long as you like.

The graph view comes with its own sign-in. You find your password for it in the dashboard under Profile.

Not automatically. knowmind stores your memories exclusively on our server in Nuremberg. As soon as you ask your AI a question, knowmind looks up the five to ten most relevant memories and sends only those as context to the AI provider you chose (Claude, ChatGPT, Gemini). The rest of your knowledge store stays with us. The audit log shows you exactly what was queried and when.

That is in the provider's domain and we cannot prevent it. If you want to rule it out, store an API key from a provider with a no-training guarantee (the Anthropic API plan, for instance, or Mistral La Plateforme with the privacy option enabled) and avoid the consumer apps from OpenAI and Anthropic for sensitive requests.

It works. The audit log records read and write operations on your knowledge store: creating, updating, searching, linking, token creation, sign-in. Pure read operations from an already signed-in browser (clicks in the dashboard) do not create an entry. What shows up is mainly MCP calls (your AI asking knowmind), API token actions, sign-in and sign-out, and configuration changes. If all you did today was sign in, you see one entry.

knowmind runs entirely in your own infrastructure on the Enterprise plan (Docker Compose or Kubernetes). On the other plans the knowledge store sits on our server, and the AI providers need a connection too. A purely offline variant for individual users is not planned — the effort for secure local data handling outweighs the benefit for most use cases.

Roughly 10 to 20 KB per memory (text plus vector index plus relations). A typical mid-sized workspace with 1,000 to 2,000 memories comes to 20 to 40 MB. You see the exact figure in the dashboard under Statistics in the storage block.

You can export your data as JSON at any time (button in the dashboard). After cancellation the workspace stays read-only for 30 days so you can make further exports. After that all data is deleted for good — Postgres, Neo4j and the vector indexes. You receive confirmation of the full deletion by email.