Airbyte, creator of the open data movement platform, announced major updates to its Airbyte Agents platform, expanding how AI agents access and act on enterprise data.

The update includes Airbyte’s availability in the OpenAI App Marketplace, the launch of the Airbyte Agent CLI, and new write capabilities for Salesforce and Hubspot that give agents the ability not just to retrieve information, but to take action.

“AI agents are quickly evolving from passive assistants into active operators,” said Michel Tricot, CEO and co-founder, Airbyte. “For agents to create real business value, they need reliable access to data and the ability to take action. With these updates, Airbyte is making it easier to build agents that don’t just answer questions, but actually get work done.”

Airbyte Now Available in the OpenAI App Marketplace

Airbyte is now available in the OpenAI App Marketplace, enabling users to connect AI agents directly to hundreds of enterprise data sources through Airbyte’s growing catalog of connectors. This gives organizations a faster path to operationalizing AI with secure, governed access to customer data, internal systems, and business applications.

By making Airbyte discoverable and deployable within the OpenAI ecosystem, organizations can rapidly integrate agentic workflows into their existing business processes without requiring custom data pipelines to be built. Support for secure OAuth and credential management is integrated, plus the built-in Airbyte Agents Context Store indexes data for fast search.

Introducing the Airbyte Agent CLI

Airbyte also introduced the Airbyte Agent CLI, a new single-binary command line interface that brings Airbyte Agents directly into developer terminals, scripts, and CI/CD pipelines.

The Airbyte Agent CLI enables developers and AI agents to:

• List available connectors;
• Inspect connector schemas;
• Execute reads, searches, creates, and updates across connected systems;
• Discover command structures dynamically through schema discovery.

This new interface gives developers and AI systems a lightweight, programmable way to work with enterprise data without leaving their existing toolchains.

The CLI provides another interface option for accessing Airbyte Agents – each meant for specific use cases. For deep integrations with Python, there is an SDK. Or, for building agentic workflows directly inside a Large Language Model (LLM) such as Claude, ChatGPT, or Cursor, support is provided for the Model Context Protocol (MCP).

Salesforce Connector Adds Write Operations

Airbyte’s Salesforce Agent Connector now supports write operations, enabling agents to update records directly inside Salesforce.

The addition of write actions transforms Airbyte from a data access layer into an operational execution layer for AI systems. Until now, many agents could only retrieve and summarize data. With write capabilities, agents can now execute tasks such as updating customer records, changing status, and automating operational workflows. For example, sales teams can make real-time changes to deal status using ChatGPT and have those updates automatically reflected in Salesforce.

Building the Infrastructure for Agentic Enterprise Workflows

These updates build on Airbyte’s broader vision of creating an agent-native data platform, one designed for both human operators and AI systems.

As enterprises adopt AI agents across sales, support, operations, and engineering, the need for structured, governed, bidirectional access to enterprise systems is becoming foundational. Airbyte’s platform provides that connective layer, enabling AI systems to securely interact with data wherever it is located.