Ships with its own skill
The install script drops an agent skill next to the binary. Your harness learns the canonical command sequence before you finish reading this page.
$
npx skills add airbytehq/airbyte-agent-cli
Airbyte Agents · Agent CLI
The CLI for everything
your company runs.
50+ business systems, one shell command away.
Built for agents first.
$
curl -fsSL https://airbyte.ai/install.sh | bash
50+ connectors and counting
Azure Blob Storage
Warehouses and Lakes
BigQuery
Warehouses and Lakes
Facebook Marketing
Marketing Analytics
Github
Warehouses and Lakes
Azure Blob Storage
Warehouses and Lakes
BigQuery
Warehouses and Lakes
Facebook Marketing
Marketing Analytics
Github
Warehouses and Lakes
Azure Blob Storage
Warehouses and Lakes
BigQuery
Warehouses and Lakes
Facebook Marketing
Marketing Analytics
Github
Warehouses and Lakes
Gitlab
Unstructured
Google Analytics
Marketing Analytics
Google Drive
Unstructured
HubSpot
Analytics
Gitlab
Unstructured
Google Analytics
Marketing Analytics
Google Drive
Unstructured
HubSpot
Analytics
Gitlab
Unstructured
Google Analytics
Marketing Analytics
Google Drive
Unstructured
HubSpot
Analytics
Built for AI agents
Your agent already knows
how to drive it.
Point Claude Code, Codex, or any shell-capable harness at airbyte-agent and it works on the first try. No tool definitions to write, no wrapper to maintain.
Claude Code
Codex- Cursor
LangChain- CrewAI
- any harness that runs commands
No guessing at flags
Run schema before any operation and get the full parameter spec without an API call. Your agent composes a valid command on the first try instead of burning turns on trial and error.
$
airbyte-agent schema connectors execute
Errors worth parsing
Failures land as typed JSON on stderr. Your agent branches on auth_error, not_found, or rate_limited instead of regexing prose.
$
{"type":"not_found","message":"connector "gong" not found in workspace "default","retryable":false}
Walkthrough
Five commands to youryour first answer.
No SDK, no dependencies, nothing to deploy.
Just type directly into your terminal.
~/airbyte-agent — quickstart
How it works
Learn it once.
Simple syntax. JSON in, JSON out. The pattern you learn on GitHub works on all 50+ connectors.
[step 01]
Connect
Create a connector once. OAuth, API keys, and token refresh are handled for you.
connectors create --json '{ ... }'[step 02]
Discover
Inspect entities, actions, and schemas before you call anything.
connectors describe --json '{ ... }'[step 03]
Ask and Act
Search and read across systems, and write back to them too.
connectors execute --json '{ ... }'platform
Same Context Store.
Four doors in.
CLI, SDK, MCP Gateway, UI: every interface searches the same pre-indexed context layer.
FAQ
Didn't find your answer?
Please don't hesitate to reach out.
What is the Airbyte Agent CLI?
The Airbyte Agent CLI (airbyte-agent) is a Go command-line interface for Airbyte Agents. Use it from a terminal, script, CI job, or AI agent harness to add connectors, inspect schemas, and run read and write actions against 50+ business systems.
How do AI agents use the Airbyte Agent CLI?
All three reach the same Context Store. The SDK is a typed Python library for custom agents, the MCP Gateway is one connection for clients like Claude and Cursor, and the CLI is the shell-first option for scripts, CI, and harnesses that run commands. It's also the non-Python alternative.
How is the CLI different from the SDK or MCP Gateway?
Run the install script (curl -fsSL https://airbyte.ai/install.sh | bash), use Homebrew (brew install airbytehq/tap/airbyte-agent-cli), or download the binary from GitHub releases. Then run airbyte-agent login.
Is it safe to put credentials in CLI commands?
You never do. Airbyte credentials live in a local settings file with 0600 permissions, and third-party credentials are entered in a browser widget so secrets stay out of shell history, logs, and agent transcripts.
How do I install the Airbyte Agent CLI?
Run the install script (curl -fsSL https://airbyte.ai/install.sh | bash), use Homebrew (brew install airbytehq/tap/airbyte-agent-cli), or download the binary from GitHub releases. Then run airbyte-agent login.
Give your agents a shell they can trust.
One line to install. Free tier, no credit card.
$ curl -fsSL https://airbyte.ai/install.sh | bash
