BotwarrenBotwarren / Integrations / Claude Code
Watch your Claude Code sessions live
The Botwarren ship agent notices when a Claude Code session starts, works, waits or stops, and turns that into a crew member on your station. Your teammates, your stream audience or your future self can watch the work happen, without ever seeing a prompt or a line of code.
Verified with real session logsAgent tests run on Linux, macOS, Windows
What the adapter looks at
The metadata of Claude Code's local session logs: file names and modification times, which project folder a session is in, the kind of tool call, whether it failed, the model family, and token counts that are reduced on your machine to a four-step context gauge. It does not keep or send your prompts, the replies, your code or file contents. The redaction test suite fails the build if planted canary text from any of those fields reaches an event or a log.
What visitors see, about 90 seconds late
- One crew member per Claude Code session: working, idle or gone.
- Coarse step hints: thinking, reading, editing, running a command.
- A stumble animation when a tool call really fails.
- Sub-agents as helpers beside their main crew member, and a rough context gauge.
- Builds you wrap with
botwarren run -- <command>as machines that run, pass or fail. Arguments and output are never sent.
Set it up
- Get a ship: log in, name it, install the agent and approve the short code.
- Start a Claude Code session inside the project folder you gave the agent. Sessions elsewhere on your computer stay private.
- Open your ship at your-name.botwarren.com, or add the OBS overlay to your stream.
Botwarren is an independent product of TurboKast Ltd. It is not made, endorsed or sponsored by Anthropic or OpenAI; Claude Code and Codex are their products.