Enkido

Up to four AI models work on one design. One is the Designer, the others are Reviewers that check its work and discuss with it. Pick models from different companies: they make different mistakes, so they catch each other's. Your keys stay on this server and are never sent back to the browser.

Enkido knows many AI services already. If yours is missing and it speaks the OpenAI chat format (almost all do), add it here and it appears in the provider list. For Azure OpenAI, Amazon Bedrock or Google Vertex AI, run a LiteLLM proxy and choose "LiteLLM proxy" as the provider.

Agent nameOrganisation (in the header and on every design) Discussion rounds (1 to 8)More rounds: more thorough, slower, costlier. Role and shared statementSent FIRST to every AI model, in every step: designing, reviewing, answering questions, reviewing the drawing. Leave it empty to use the default. Parallel cablesFour cables between the same pair are drawn as one line marked "4 x 25G". Every cable is still listed in the documentation and the Bill of Materials. Review the drawing visuallyA reviewer that can see images looks at the finished drawing and suggests a tidier layout. The design itself never changes.

Reading online. Claude and Gemini search by themselves. A LOCAL model (Ollama, LM Studio, vLLM) cannot: the local server does not browse, and small models never ask for a search. For those, Enkido does the reading itself: it asks the model what to look up, searches, opens the pages and puts the text in front of it. That needs a search service below, and "Reads online" on the model.

Search service SearXNG addressSearch API key

Enkido

Conversation

Settings

Sign in to Enkido

Account