Skip to main content
The Corbado Observe CLI gives you scripted access to your Observe data from the terminal, without going through the management console. It is built for humans and for AI coding agents alike: every command explains itself through --help, output is JSON first and the full command schema is available for an agent to read.

​
1. What it is for

​
2. Install

The CLI is available on Apple Silicon macOS and glibc-based Linux x64 and arm64. Windows is available on request through support.

​
macOS

On macOS, install it with Homebrew from the Corbado tap:
Homebrew puts corbado on your PATH. The CLI is a signed and notarized standalone executable, so it runs without a Node.js or Bun installation. Upgrade with brew update followed by brew upgrade --cask corbado/tap/corbado. Standalone downloads and release notes are on the releases page.

​
Linux

Install the latest release on a Linux x64 or arm64 machine:
The installer detects your architecture, verifies the release checksum and installs corbado in ~/.local/bin. Follow its PATH instruction if needed. No Bun, Node.js or root access is required. The CLI works per project. On macOS, add the first one with:
This opens the management console at app.corbado.com, where you log in and pick the project to link. The CLI stores project metadata in a config file and the API key in the operating system’s secret store. On Linux VMs or agent environments without a secret store, provide an Observe API key and project through environment variables instead:
Provide the key through your VM’s secret management rather than saving it in a shell profile. This path does not require libsecret or interactive pairing. When you are unsure which permissions the link needs, start with read permissions. Every analysis and export works with read access, and you can raise the permissions later under Observe → Settings → API keys. Switch between linked projects with corbado switch. For parallel agents using stored credentials, set CORBADO_PROJECT_ID per process or pass --project-id on each call. Agents using environment credentials supply the matching project ID with each key.

​
4. First steps

corbado guide is the investigation playbook. It has topics such as errors, experiments, alerting, exports and history, for example corbado guide errors. corbado config returns the catalog of your project’s data: series, dimensions and how the entities connect. corbado schema prints the machine-readable command documentation, which is what a coding agent reads to drive the CLI on its own.

​
5. Use with coding agents

The CLI is optimized for every coding agent: Claude Code, Cursor, Codex, Copilot and others. Give the agent the task in plain words and name the CLI in the prompt, because coding agents do not know it from their training yet:
The CLI explains itself to the agent, so no extra skill or setup is needed: --help and corbado schema describe the commands, corbado guide gives the investigation playbook and corbado config describes how your project’s data is structured.

​
6. Design

The CLI follows the gh model: typed commands for the common path, a raw API escape hatch for everything else, JSON-first output that can be filtered inline and full schema introspection.

Instrument with AI agents

Add custom events to your application with the agent skill.

MCP server

Observe data for agents through the Model Context Protocol, in closed beta.