Octo - Minglue Technology's open-source AI-native team collaboration platform

What is Octo

Octo is Minglue Technology's open-source AI-native team collaboration platform. It aggregates scattered AI Agents into a unified space and enables efficient orchestration between humans and Agents and between Agents themselves through an IM format. Its core includes Matter, Taste, and six collaboration modes, including Solo, Roundtable, Critic, Pipeline, Split, and Swarm. It supports private deployment and is positioned as enterprise organizational infrastructure for the Private AI era, upgrading Agents from personal tools to shareable digital assets.

Octo's main features

• Bot aggregation and sharing: Centralize Bots from various workflows into the same space, support personal use and authorized sharing within the organization, and upgrade Bots from personal tools to enterprise-level digital assets.
• Matter: Consolidate complex tasks into traceable "decision cards," including the task brief, process timeline, key outputs, feedback, and acceptance, forming organizational memory.
• Taste (preference evolution): Consolidate implicit feedback such as human rejections, modifications, and confirmations in real-world practice into reusable preferences for Agents, achieving "the more you use it, the better it understands you."
• Six collaboration modes: Solo, Roundtable, Critic (generate-verify), Pipeline, Split, and Swarm, matching different information topologies.
• Structured space: A four-layer architecture of Space → Category → Channel → Thread organizes collaboration relationships clearly and prevents messages from being drowned out.
• Voice input: Supports context-aware voice transcription and correction, allows voice @mentions and content modification, and improves input efficiency.
• Browser extension (Cmd+K): Select webpage content and bring it into context with one click. Bots take over immediately without switching tool flows.

Follow WeChat and reply "open source" to jointhe AI open-source project exchange group

How to use Octo

• Visit the official website/repository: Go to the GitHub repository to get the open-source code.
• Private deployment: Complete deployment on a local server or private cloud environment according to the documentation.
• Create a Space: Establish a goal-oriented workspace and invite members and Agents to join.
• Adopt Bot templates: Choose Bots from ready-made templates, or connect runtimes such as OpenClaw, Codex, and Claude Code.
• Initiate a Matter: Create an item in a Channel/Thread, set the Brief, and choose a collaboration mode (such as Pipeline or Critic).
• Advance through human-AI collaboration: Agents automatically decompose, schedule, and execute, while humans make judgments and trade-offs at key nodes.
• Accumulate Taste: Through feedback, rejections, confirmations, and other operations, let Agents learn organizational preferences.

Octo's core advantages

• Organization-level Agent network: Break down standalone Agent silos and enable a Bot army to flow freely and collaborate with division of labor within the organization.
• Data sovereignty and privacy: Follow a private deployment path, with open-source white-box auditability, ensuring that enterprise context and judgment standards do not leak.
• Accumulation of practical preferences: Transform humans' tacit judgment (Taste) into structured preferences that Agents can inherit, rather than simply distilling them into platform assets.
• Six information topologies: Compared with IM group chats, it supports refined collaboration where visibility is shown when it should be visible and hidden when it should be hidden.
• Zero-threshold use: Ready-made Bot templates are provided, requiring no cumbersome configuration; adopt one and pull it into the group to get work done.

Octo's project address

• GitHub repository：https://github.com/Mininglamp-OSS

Comparison of Octo with similar competitors

Octo's application scenarios

• Complex enterprise project management: Use Matter to accumulate the decision-making process, with multiple Agents collaborating in different roles to complete long-horizon tasks.
• Code review and quality inspection: In Critic mode, one Agent generates code, and another independently reviews it; if it fails, it is sent back for rework.
• Industry research report writing: Split mode divides the report into modules such as policy, market, and technology, with multiple Agents working in parallel and dividing the work.
• Creative proposal competition: Swarm mode assigns the same task to multiple Agents to produce independently, and the Leader selects the best.
• Accumulation of private AI assets: Local deployment ensures data sovereignty and transforms the organization's collaboration process into long-term competitiveness that cannot be replicated.