openTag features explained
openTag feature overview
openTag is an open source AI coworker for Slack and Lark. Its basic pattern is chat-first: you mention openTag in a thread, hand it a task, and it sends that work to an agent running on your own computer before replying in the same conversation. That keeps the request, the execution, and the answer close to the discussion where the work started.
A central feature is task delegation inside messaging channels. openTag is designed so one agent can serve a channel, letting several people in that workspace mention it when they need help. It also supports agents handing work to each other and checking on progress, which makes it useful for work that moves through more than one step rather than stopping at a single reply.
openTag also runs locally. The agent works on the user’s Mac or Linux machine, and openTag says messages and results pass through the product while the actual work happens on that computer. For teams that want the assistant tied to their own setup, that local execution model is a defining part of how the product works.
Provider choice is another major feature. openTag is built without AI provider lock-in, so users can bring their own subscription or API key. The site says it works with Claude, Codex, and Pi on the machine that runs the agent. That matters because openTag sits above the model choice instead of replacing it with a separate bundled assistant account.
Shared team memory is handled through Context Tree. openTag describes it as a portable, file-based, git-backed memory layer hosted in the team’s GitHub organization. In practical terms, that gives the agent a shared record of notes, decisions, goals, and patterns that can be reused later instead of living only in a single chat history. Because the memory is stored as files in a repository the team hosts, it stays readable and versioned rather than hidden inside a separate database.
openTag also connects to a broad set of external services. The product says its agents connect to more than 1,000 popular services, including GitHub, Notion, Google Drive, Jira, Figma, and Linear. That makes the agent more useful when a task depends on information outside Slack or Lark, such as project notes, documents, tickets, or design files.
The control model ties these features together. openTag says it can work with your own model subscription or API key, and by default its servers do not store conversations, code, or memory. The result is a product built around chat delegation, local execution, shared memory, and external connections, all centered on the user’s own computer and workspace.