Octo - An open-source AI-native team collaboration platform from Minglue Technology

What is Octo?

Octo is an open-source AI-native team collaboration platform from Minglue Technology. It aggregates scattered AI agents into a unified space, enabling efficient orchestration of human-agent and agent-agent interactions through an instant messaging (IM) mechanism. Its core features include Matter, Taste, and six collaboration modes: Solo, Roundtable, Critic, Pipeline, Split, and Swarm. Supporting private deployment, it is positioned as an enterprise organizational infrastructure for the Private AI era, upgrading agents from personal tools to shareable digital assets.

Octo's main functions

• Bot aggregation and sharingThis feature centralizes bots from various workflows into a single space, supporting both personal use and authorized sharing within an organization, thus upgrading bots from personal tools to enterprise-level digital assets.
• MatterComplex tasks are distilled into traceable "decision cards," which include the task brief, timeline, key outputs, feedback, and acceptance, forming organizational memory.
• Taste (Preference Evolution)The implicit feedback from humans in real-world scenarios, such as rejections, modifications, and confirmations, is distilled into reusable preferences for the Agent, enabling it to "understand you better the more you use it."
• Six Collaboration ModesSolo, Roundtable, Critic, Pipeline, Split, and Swarm are used to match different information topologies.
• Structured spaceThe four-layer architecture of Space → Category → Channel → Thread clearly organizes the collaboration relationships and prevents messages from being overwhelmed.
• Voice inputIt supports context-aware speech transcription and correction, allowing users to @ others and modify content via voice, thus improving input efficiency.
• Browser plugin (Cmd+K)Select webpage content and import it into the context with one click; the bot will immediately take over the processing without switching tool flows.

Follow us on WeChat and reply with "open source",join inAI open source project discussion group

How to use Octo

• Visit the official website/repositoryAccess the GitHub repository to obtain the open-source code.
• Private DeploymentDeployment can be completed on a local server or in a private cloud environment, according to the documentation.
• Create SpaceEstablish a goal-oriented workspace and invite members and agents to join.
• Adopt a Bot TemplateChoose a Bot from existing templates, or integrate with runtimes such as OpenClaw, Codex, and Claude Code.
• Initiate MatterCreate a task in Channel/Thread, set a Brief, and select a collaboration mode (such as Pipeline or Critic).
• Human-machine collaborative advancementThe agent automatically breaks down, schedules, and executes tasks, while humans make judgments and choices at key nodes.
• Precipitation Taste: By providing feedback, rejection, and confirmation, the Agent learns organizational preferences.

Octo's core advantages

• Organizational Agent NetworkBreak down isolated agent silos and enable the free flow and collaborative division of labor among bot armies within the organization.
• Data sovereignty and privacyBy pursuing a privatization path, while maintaining open-source white-box auditability, we ensure that the company's context and judgment criteria are not leaked.
• Practical preference accumulationThe goal is to transform implicit human judgments (Taste) into structured preferences that can be inherited by the Agent, rather than simply distilling them into platform assets.
• Six Information TopologiesCompared to IM group chat, it supports refined collaboration that allows users to see each other when appropriate and to remain anonymous when necessary.
• Zero-barrier to useWe provide ready-made bot templates, requiring no complicated configuration; simply adopt a bot and bring it into the group to work.

Octo's project address

• GitHub repositoryhttps://github.com/Mininglamp-OSS

Octo's Competitive Product Comparison

Application scenarios of Octo

• Enterprise complex project managementMatter helps to solidify the decision-making process, enabling multiple agents to collaborate in different roles to complete long-term tasks.
• Code review and quality inspectionIn Critic mode, one agent generates the code, another reviews it independently, and if it fails, it is sent back for redo.
• Industry research report writingThe Split mode breaks down the report into modules such as policy, market, and technology, with multiple agents working in parallel.
• Creative proposal competitionThe Swarm mode assigns the same task to multiple agents to produce independently, and the leader selects the best one.
• Privatization of AI AssetsLocal deployment safeguards data sovereignty and transforms organizational collaboration processes into a unique, long-term competitive advantage.