CompNav · Open-source projects
Super Agent Party
A self-hosted AI desktop companion and VTuber workspace that combines chat, VRM/Live2D characters, messaging integrations, and livestream interaction.
What it is
Super Agent Party (SAP) combines AI conversation, desktop characters, agent tools, and external integrations in a desktop app and self-hostable service. Users can chat with characters, show VRM avatars on the desktop, or control Live2D through VTube Studio. Characters can also connect to messaging platforms such as QQ, WeChat, Feishu, Telegram, and Discord, or support interaction on Bilibili, YouTube, and Twitch streams. It includes many general agent and automation features, making it an extensible character workspace rather than a simple chat companion.
Core mechanisms and capabilities
The desktop companion supports custom VRM characters, motions, and scenes. For Live2D, the project integrates with VTube Studio so voice or interaction can drive expressions and actions.
The chat interface supports character personas, Tavern character cards, multi-character group chat, and long-term memory; characters can use different avatars and voices. This supports both everyday conversation and interaction centered on character relationships and scenarios.
The project connects to several messaging and livestream platforms and provides an extension system. These integrations let characters appear in existing communities or streams, but each platform has its own accounts, credentials, and configuration requirements.
Beyond companionship, SAP offers agent capabilities such as MCP, Agent Skills, browser access, and desktop control. These are general workspace features and should be configured with care; they are not required for character chat itself.
Editorial take
Super Agent Party is included because it packages AI companion characters, desktop avatars, and VTuber/livestream workflows into a downloadable, self-hostable application, with an AGPL-3.0 repository and a visible community. Its breadth is its defining trait: alongside character chat it includes messaging bots, livestream connections, extensions, and computer control. That breadth also adds complexity—models, voice services, and external platforms still require configuration, and users seeking a lightweight chat app may not need most of the workspace features.
Who it suits
Suited to people who want AI characters on the desktop, in livestreams, or on messaging platforms, especially when VRM/Live2D, group chat, voice, or extensions matter. It also fits users who want character interaction and agent tools in one self-hosted environment.
Deployment and technical requirements
Official distribution includes a Windows installer and portable package, an Apple Silicon macOS build, Linux AppImage, and Docker deployment. The desktop app can host character windows; the Docker version is primarily browser-based, with a separate lightweight desktop client documented for it. Model providers, voice features, and external integrations require additional in-app configuration.
Runtime requirements
Models can use hosted services or a local deployment, so provider and model configuration is required. Speech synthesis, speech recognition, and messaging or livestream integrations may each need their own API keys, accounts, or platform credentials. The project reports testing its Docker server mode with 2 CPU cores and 2 GB RAM; that does not imply the same requirements for desktop avatars, voice, or local models.
The project does not require one specific commercial model and can connect to hosted services or local models; users must configure the provider, endpoint, and credentials in the app. Speech synthesis, speech recognition, and platform bots may need separate service credentials or components. The project reports that its Docker server mode can run on a modest server, but local inference and desktop multimedia needs depend on the chosen model and avatar assets.
Limitations and considerations
The main project uses AGPL-3.0; modifications offered as a network service may trigger source-code obligations under the license. The README also notes that some voice features depend on public or experimental third-party interfaces and may stop working if policies change. Avatar models, extensions, and external platforms can have separate licenses or service restrictions.
Official sources and further reading
Information checked: Oct 26, 2026.