Get going with NomiFun
A curated set of operational, how-to guides for getting this local-first super AI workstation up and running. For deep technical internals and architecture, see the full docs on GitHub.
Getting Started
- Introduction Meet NomiFun: a cross-platform AI workspace with conversations, collaboration, Skills, presets, speech input, automation, and desktop companions.
- Installation Install NomiFun on Windows, macOS, or Linux, or build from source and self-host with Docker.
- Your first conversation Configure a model, start your first session from the home screen, and use the workspace's messages, file tree, preview, and terminal.
Models & Decisioning
- Model routing & failover Connect the 4 native provider backends and orchestrate a failover queue of up to 4 switches so a session keeps going when a provider fails or is rate-limited.
- Intelligent decisioning (IDMM) A per-session supervisor — a rule layer handles mechanical stalls and a side-model layer resolves real decision checkpoints, so unattended work reaches a terminal state.
Sessions & Terminal
- The session workspace A session = its own working directory + message stream + file tree + multi-format preview + a terminal mounted to that directory, with knowledge bases, AutoWork, and decision watch in the header.
- Built-in terminal A native in-app PTY terminal with Shell / Claude Code / Codex / Gemini presets — attach it to a session's working directory and drive it by hand or hand it to AutoWork.
Companions & Channels
- Companions Raise a family of companions — shared/private memory scopes, per-companion skill libraries, knowledge-base binding, skill evolution, and cross-machine migration.
- Channels Connect a companion to 12 built-in IM channels, authorize chat users with a QR scan or pairing code, and command your computer from a phone or group chat.
Knowledge & Open Capabilities
- Knowledge bases Manage local Markdown, URL snapshots, and write-back policy so any Agent can share traceable context across sessions, terminals, and companions.
- MCP & Skills Connect external MCP servers, manage installed Skills, and discover and review new capabilities in Skill Market.
- Open capabilities (MCP / REST / OpenAPI) Expose NomiFun's platform capabilities to any agent through /mcp, /mcp-agent, and /v1 with a single companion-token, and delegate tasks as that companion.
Knowledge & Open Capability
Automation
- Requirements & AutoWork Register requirements, bind a tag to a session or terminal, and let the orchestrator run them unattended — with outbound completion notifications.
- Scheduled tasks Schedule companions with cron, run a job on demand, review run history, and keep the system awake.
Remote & Settings
- WebUI remote Expose the desktop backend over LAN behind a login; scan a QR with a 5-minute one-time token to work from your phone or tablet, with the same UI as desktop.
- Speech input Configure cloud or local speech recognition in Model Management, then record directly from the conversation composer.
- System settings Adjust themes, language, startup and notification preferences; manage execution engines, Browser Use, Computer Use, and application updates.
- Computer-use & browser Let an agent operate your desktop and browser, with clear browser visibility, login state, system permissions, and action approvals.
Creation & Mini Apps
- Agent Desktop mini-apps Turn ordinary conversations into single-file web tools that you can publish, run, iterate, or import locally.
- Creative Workshop (Beta) Arrange images, text, video, and generation nodes on an infinite canvas for local-first multimedia creation.
Services & Customer Care
Devices & Ecosystem
- Connect a XiaoZhi robot Let a compatible ESP32 robot use NomiFun companions, models, voice services, and device-side MCP tools over the LAN.
- Mobile bridge to Desktop Use NomiFun Mobile to inspect tasks, send instructions, and handle confirmations while connecting back to your own Desktop.