Octopus
An open-source multi-agent collaboration tool developed by Tencent. It supports multi-agent expert team scheduling, a Connect connector framework, long-term memory and knowledge bases, cross-platform collaboration, and ACP-based connections to external agents. The tool is designed for developer-oriented users and small team scenarios.
Timeline (last 90 days)
Tencent has open-sourced an Agent tool called Octopus.
Octopus has accumulated more than 7,000 stars on GitHub.
Octopus is more akin to an upgraded encapsulation of OpenCode, targeting geek users who enjoy tinkering, rather than being a consumer-grade Personal Agent.
Octopus's Connect connector system can integrate with the Tencent ecosystem and supports login and connection methods for WeCom, DingTalk, Feishu, and others.
Octopus is equipped with a long-term memory module and a knowledge base, and its knowledge base vector model supports local installation and deployment.
Octopus supports collaborative work across Windows, Linux, and macOS, and allows real-time viewing of mouse operations in the console.
Octopus uses SQLite as its database, making it not well-suited for genuine multi-user, high-concurrency team use.
Octopus has a built-in skill marketplace, supports skill recording and playback, and can record manual operations for automatic replay and execution.
Octopus supports remote launching of sandbox environments, where the AI works inside the sandbox and can operate remote desktops (such as Linux servers).
Octopus can connect to various locally installed agents via ACP, acting as the brain to schedule and control more agents.
1 more timeline events
All Facts (10)
In actual testing, Octopus can be connected to the DeepSeek model, which requires configuring the model API Key.
50%UnverifiedOctopus is more akin to an upgraded encapsulation of OpenCode, targeting geek users who enjoy tinkering, rather than being a consumer-grade Personal Agent.
50%UnverifiedOctopus's Connect connector system can integrate with the Tencent ecosystem and supports login and connection methods for WeCom, DingTalk, Feishu, and others.
50%UnverifiedOctopus is equipped with a long-term memory module and a knowledge base, and its knowledge base vector model supports local installation and deployment.
50%UnverifiedOctopus supports collaborative work across Windows, Linux, and macOS, and allows real-time viewing of mouse operations in the console.
50%UnverifiedOctopus uses SQLite as its database, making it not well-suited for genuine multi-user, high-concurrency team use.
50%UnverifiedOctopus has a built-in skill marketplace, supports skill recording and playback, and can record manual operations for automatic replay and execution.
50%UnverifiedOctopus supports remote launching of sandbox environments, where the AI works inside the sandbox and can operate remote desktops (such as Linux servers).
50%UnverifiedOctopus can connect to various locally installed agents via ACP, acting as the brain to schedule and control more agents.
50%UnverifiedTencent has open-sourced an Agent tool called Octopus.
50%