Airbyte, creator of the open data movement platform, today announced major enhancements to the Airbyte Agents platform, making it easier for AI agents to discover relevant enterprise knowledge while ensuring organizations maintain precise control over what agents and users can access.
As enterprises move from AI experimentation to production, two challenges consistently emerge: giving agents access to the vast amount of unstructured business knowledge spread across collaboration tools, and ensuring that access is governed by security policies. Airbyte’s semantic search support and new entity policies for workspaces address both challenges through meaning-based retrieval and fine-grained governance built directly into the Context Store – a replicated, search-optimized index – that is part of the Airbyte Agents platform.
“AI agents are only as valuable as the context they can safely access,” said Michel Tricot, CEO and co-founder of Airbyte. “Organizations don’t need another disconnected vector database or another permission system, they need agents that understand the information that already exists across their business while respecting the same governance policies employees rely on every day. These new capabilities move us another step closer to making enterprise AI both more useful and more trustworthy.”
Semantic Search Brings Meaning-Based Retrieval to AI Agents
Airbyte Agents now supports semantic search across content stored in Google Drive, Gong call transcripts, Granola meeting notes and Linear issues and comments, which cover the main places where organizational knowledge exists outside of structured records.
This enables AI agents to retrieve information based on meaning rather than exact keywords. Instead of relying on literal text matches, semantic search understands intent and context. An agent can identify customer conversations discussing pricing concerns, contract objections, competitive products, login failures, or engineering issues – even when those exact terms never appear.
Rather than searching for specific file names or phrases, agents can locate relevant information buried within documents, meeting notes, and collaborative content – even when the language used differs from a user’s query. This allows organizations to ask natural-language questions such as:
- What customer concerns were discussed during last week’s planning meetings?
- Where have we discussed pricing strategy across documents and meeting notes?
Because semantic search operates on Airbyte’s pre-indexed Context Store rather than repeatedly querying source APIs, organizations benefit from significantly lower inference costs and faster response times. Internal benchmarks have demonstrated up to 80% fewer tokens when querying Gong and up to 75% fewer tokens for Linear compared to native API approaches.
Additional data connectors will receive semantic search support in future releases.
New Entity Policies Deliver Fine-Grained Governance
Last month, Airbyte Agents added a workspaces feature that gives users their own separate instance for Airbyte Agents. Each workspace can have its own users and defined access to specific data connectors. Now, a new Airbyte Agents feature introduces entity policies for workspaces, which provide precise control over which data connectors, data sources, and other resources, can be discovered and accessed by users and AI agents.
Organizations can now define policies that determine visibility and access at the entity level (data connector, data sources), allowing teams to expose only the information appropriate for specific departments, projects, or users. This enables agents to be deployed at scale with security controls. Now, read and write policies can be assigned to every user and agent in every data connector, across every workspace. This enables data connectors to be shared among users and agents while keeping sensitive information restricted to those with appropriate permissions.
This enables enterprises to:
- Restrict agent access to data sources;
- Separate development, staging, and production environments;
- Limit visibility of sensitive business systems;
- Align AI access with existing organizational security and compliance requirements.
The new governance capabilities help enterprises confidently deploy AI agents without sacrificing operational control or creating duplicate permission models. Entity policies provide fine-grained control over what users and agents have access and write capabilities.
By combining semantic retrieval with enterprise-grade governance, organizations can build AI agents that not only retrieve more relevant information, but do so within clearly defined boundaries.
Building the Enterprise Context Layer for AI Agents
These capabilities advance Airbyte’s vision of creating the industry’s leading data platform for agentic AI.
Airbyte provides the foundation for production AI systems by combining enterprise connectivity, indexed contextual understanding, semantic retrieval, governance, and operational execution into a unified platform for AI agents.
Go here for a free trial of Airbyte Agents.
About Airbyte
Airbyte is the context infrastructure platform for AI agents and analytics. Airbyte Agents gives every agent a unified view of operational data through a replicated, search-optimized Context Store, a native SDK, an MCP endpoint that works with any MCP-compatible client, and a CLI. Airbyte Agents is built on Airbyte’s open-source data replication platform, powered by the industry’s largest ecosystem of connectors and trusted by 7,000 enterprises to move structured and unstructured data across multi-cloud and hybrid environments. Data and AI teams use Airbyte to build both pipelines and agents on a shared foundation. For more information, visit https://airbyte.com.
View source version on businesswire.com: https://www.businesswire.com/news/home/20260825317819/en/
Media gallery
