The CLI is the command-line version of Landbase that runs inside an AI assistant such as Claude Code or Codex. You get the same data and the same agent from your terminal, where Landbase sits right next to every other tool you have connected. That placement is the point: instead of building one list at a time in the web app, you can set up a series of steps once, where the agent searches, adds detail, and hands the result to the tools where you work, and let it run for you. It is where Landbase stops being one more place to build lists and becomes part of the work itself.
Because the CLI runs in your assistant, the agent can call Landbase to find accounts and add more detail to them, then pass what it finds to the other tools you have connected there, without you copying anything between apps. Work that once spanned several tabs now happens in one conversation. A quick way to confirm the setup is to run a single search and watch the results come back in your terminal.
Each tool connects through its own connection inside your assistant. A CRM such as HubSpot or Salesforce links this way, with a one-time secure sign-in, and today only the CLI can do this. The web app cannot connect a CRM yet. Reading lets Landbase see your records, so it can remove duplicates from a new list by comparing it to what you already have, or build lookalike lists from the deals you have already won. Writing lets Landbase add to your records, so a finished list with all the detail lands in your CRM without an export. Start with read-only access and confirm a small pull before you let the agent write.
A real task rarely stops at one step. Building a list, adding more detail to it, matching it against your CRM, and writing the new records back are four actions in a row. Connecting steps so they run one after another lets you describe the whole series once, and the agent runs each step in order, passing the result of one into the next. Build it one step at a time and confirm each one before you run the whole thing start to finish.
Once a series of steps is built and tested, you hand the whole thing to the agent. It runs each step, pauses to ask whenever a step needs input or approval, and returns a summary you can check. Permissions decide what it may do on its own. Auto-approve the read-only and low-risk commands you run all the time, and keep the few actions that change data outside Landbase, such as writing to your CRM, behind manual approval. Start tighter and loosen up as a series of steps earns your confidence.
Ask your assistant in natural language.
Use the Landbase CLI to build a list of fintech startups in NYC with 50 to 200 employees, then narrow it to companies founded after 2020.Build a list of heads of marketing at B2B SaaS companies in California, enrich it with title and company size, then add the new contacts to my CRM.Pull my closed-won accounts from my CRM, use them as seeds to find lookalike companies, enrich the matches, and add the new ones back.