How to use openTag effectively
openTag usage guide
openTag is built for chat-first work in Slack or Lark. You mention it in a thread, hand off a task, and the agent runs on your own computer before replying in the same conversation. That makes the product feel like a coworker inside the workspace rather than a separate tool you have to check somewhere else.
The basic setup has three parts: create an agent, prepare the computer that will run it, and connect the messaging app. openTag supports Slack and Lark, and the supported machine side is Mac or Linux. The product also says you can use Claude, Codex, or Pi on that machine, with your own subscription or an API key.
Once the pieces are connected, the day-to-day pattern is simple. A channel member mentions openTag in a thread, the request is passed to the agent on the connected computer, and the result comes back to that same thread. openTag also says one agent can serve a channel, so the same agent can be used by multiple people in that workspace.
For teams, the other important part is shared context. openTag includes Context Tree, a file-based, git-backed memory layer hosted in the team’s GitHub organization. That memory is described as readable, portable, and versioned, which suits teams that want decisions and notes to stay available for later tasks instead of living only in chat history.
openTag also connects to more than 1,000 popular services, including GitHub, Notion, Google Drive, Jira, and Figma. In use, that means the agent can work with context beyond the thread where the request started, while still keeping Slack or Lark as the place where people ask for work and receive the reply.
The control model is part of the usage story too. openTag says the agent runs on your computer and talks to your model provider directly. By default, its servers do not store conversations, code, or memory. For shared workspaces, the site recommends using an API key.