> ## Documentation Index > Fetch the complete documentation index at: https://docs.gobihq.com/llms.txt > Use this file to discover all available pages before exploring further. # Use Gobi with AI agents > Let the AI tools you already run read your captures and act on your behalf. Your captures don't have to stay in the app. Gobi ships a command-line interface, and any AI agent that can run shell commands can use it to work with what you've captured. That covers Claude Code, or a script of your own. ## What an agent can do An agent signs in as you, so it reaches what you reach and stops where you stop. * **Read your captures.** List your conversations, pull a transcript, get the audio. In a shared space it can browse what happened, but transcripts and audio stay with whoever recorded them, agent or not. * **Write to your spaces.** Post updates, reply in threads, react. A daily summary posted to your team's channel is a few commands. * **Search.** Find that conversation where pricing came up, across everything you've captured. * **Work with artifacts.** Write markdown or upload an image or video to make an [artifact](/introduction/glossary#artifact), then attach it to a post to share it. Every edit keeps the version before it. So "summarize what I promised people this week and post it to my personal feed" is a real workflow, with your captures as the source. ## Set it up Install the CLI and sign in once: ```sh theme={null} brew tap gobi-ai/tap && brew install gobi gobi auth login ``` If your agent is Claude Code, you're done at this point: the CLI package ships skills that teach it every command group. For anything else, point it at the [developer docs](/cli/overview); every command returns structured JSON when asked. An agent using the CLI is signed in as you. It sees what you see and nothing more, and that line holds at the API: another member's transcript is refused no matter who's asking. Details in [where your captures go](/privacy/your-captures).