Chat apps, agent builders, workflow tools, and tool servers usually live in four separate products, and keinsaas Navigator folds them into one self-hosted AI workspace. Point it at Ollama, LM Studio, or any OpenAI-compatible endpoint, register the models you want, and pick one for each conversation. Type an at sign in the composer to pull an agent, a workflow, or a tool straight into the message. Agents carry their own role, instructions, model, and tool list, and each one stays private or is shared with the team. The workflow editor connects Input, LLM, Condition, Tool, HTTP, Template, Note, and Output nodes on a canvas, and a published workflow becomes a tool that any chat can call. MCP servers connect over HTTP or local commands, and a built-in tester runs each tool with JSON input before an agent depends on it. Projects keep shared documents and memory with vector search. Agents can run on a schedule for recurring jobs like morning briefs. Teams get accounts and sharing, and the first person to sign up becomes the administrator. A support desk copilot, a private assistant, or a prototype for the next automation all start from the same screen. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Libre WebUI is a self-hosted AI workspace that keeps chat, notes, personas, and agent tasks in one place you control. Local models run through Ollama, and hosted ones come from OpenAI, Anthropic, Gemini, Groq, Mistral, OpenRouter, or any OpenAI-compatible endpoint added on the provider settings page. Pick a model per conversation. Replies stream, prompts queue, chats fork into branches, and several models can answer side by side. Notes keep Markdown documents with revision history. Personas bundle a system prompt, a model, and memory into a reusable assistant, and prompt templates with typed variables appear when you type a slash in the composer. Knowledge collections ground answers in uploaded documents and cite their sources. The Work area gives each task its own workspace with files, a terminal, diffs, and a live preview, and automations run scheduled prompts that arrive as normal chats. Administrators create users and groups, hand out roles, issue scoped API tokens, and share chats, notes, and personas through one grant model. A private team assistant, a research desk, or a sandbox where agents build small apps all fit here. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.