Error: Could not load results.
The Ideenatlas provides a public, documented interface (API). It is specifically optimized to be used by both traditional applications and autonomously by AI agents (Large Language Models).
The primary search interface provides structured JSON responses by default, or token-optimized Markdown by appending '?format=md'. It includes direct search results, hierarchical paths within the knowledge graph, thematically similar clusters, and calculated serendipitous connections.
Alongside the classical API, an MCP server is also provided. Using the 'search' tool, the results are returned as a highly optimized Markdown document, making it extremely token-efficient for LLMs. No authentication is needed.
For AI assistants in 'browsing mode' (e.g., ChatGPT Web, Claude, Perplexity) that cannot send POST requests, a GET wrapper is available. It performs the vector search and formats the response into a Markdown document wrapped in a minimal, easily machine-readable HTML structure.
For AI agents that cannot construct URLs manually, but are allowed to interact with webpages, a minimalist HTML entry point is available. It contains a search form that redirects to the API view.
To ensure autonomous use by AI systems without human intervention, the Ideenatlas fully supports common machine discovery standards:
OpenAPI Spec (JSON): /v3/api-docs/public-search-api
Swagger UI (Interactive): /swagger-ui.html
LLM Instructions: /llms.txt
AI Plugin Manifest: /.well-known/ai-plugin.json
The API is open and currently does not require an API key. We ask you to refrain from massive automated scraping of the database and to adhere to rate limits of at most one request per second. Abusive use will lead to a temporary block of the calling IP address.