Switch from Claude to openTag
Switching from Claude to openTag: What Changes
Moving work from Claude to openTag does not necessarily mean leaving Claude behind. openTag can use your existing Claude plan to run a coding agent on your computer, with requests and replies handled through Slack or Lark. The main change is where you delegate work and how the agent runs—not necessarily which AI provider you use.
This is a useful distinction if your goal is to bring AI tasks into team conversations. It is a different decision from replacing Claude’s web, desktop, or mobile experience.
From a Claude conversation to a shared channel
Claude provides chat on web, desktop, and mobile, along with web search, file creation, code execution, memory across conversations, and connections to apps and tools. Its Pro plan adds capabilities including Projects and Claude Code.
openTag centers work on a thread in Slack or Lark. You mention the agent with a task, openTag passes the request to your computer, and a background program runs your coding agent. The reply returns to the same thread. One agent can serve a channel, and anyone in that channel can mention it.
For someone moving existing work, this changes the starting point: instead of opening a separate Claude conversation, you can delegate from the discussion where the task arose. For example, a team could ask an agent to work on a task already being discussed in a channel and receive its response there. That is a reason to move selected tasks to openTag, rather than treating the move as a wholesale replacement for every use of Claude.
What you need to carry the work over
openTag requires a Mac or Linux computer with Claude, Codex, or Pi installed, plus a Slack or Lark workspace. Its setup connects an agent, the computer that runs it, and the messaging workspace. This introduces a local execution environment into the way you use AI.
You can keep Claude as the underlying agent choice or use another supported option. openTag accepts a Claude or Codex subscription, or an API key. Subscription use remains subject to the provider’s terms, and openTag recommends an API key for shared workspaces.
openTag itself is free to use, but AI usage is paid for through your provider subscription or API access. Any provider charges remain separate from openTag’s free software. Claude also has a free plan; its Pro plan is listed at $20 monthly or $200 billed annually, excluding applicable tax.
Team memory is a separate transition
Claude offers memory across conversations. openTag’s shared-memory approach is Context Tree: readable Markdown files in a git repository hosted in your own GitHub organization. Multiple agents can share this memory, and git provides version history.
For a team moving work into openTag, this creates a place for reusable context such as goals, decisions, and working patterns. It is a different form of continuity from conversational memory. Important project context can be maintained as team-owned files, rather than treating a new chat interface as sufficient preparation for the move.
The execution and data paths also matter. openTag’s agents run on your computer and communicate directly with your model provider. By default, openTag’s servers do not store conversations, code, or memory, although messages and results pass through openTag. Connectors use your credentials, while Context Tree stays in the repository you host. Local agent execution does not mean the model runs offline.
A switch can be selective
openTag fits a move toward shared, chat-based delegation using a locally running coding agent. Claude remains relevant for work you want to keep in its own apps, including its file, search, and conversation-based capabilities.
You can therefore adopt openTag for channel-based tasks while continuing to use Claude directly elsewhere—and, where provider terms permit, keep Claude powering the agent behind those tasks. The practical transition is choosing which work belongs in a shared thread and supplying the local agent and team context needed to carry it out.