Michel, co‑founder and CEO of Airbyte, unveiled Airbyte Agents, a unified data layer that lets AI agents discover information and act across operational systems such as Slack, Salesforce or Linear. A short walkthrough video demonstrates how the service indexes data from the company’s six‑year‑old catalog of connectors, exposing a single searchable surface for downstream tools.

Typical middleware platforms expose thin API wrappers, forcing agents to juggle authentication, pagination and schema mapping while still missing cross‑system entities. Airbyte’s answer is the Context Store, an index optimized for agentic search that pre‑populates from its replication connectors. This gives agents a structured discovery path yet retains the ability to read or write directly to the source.

Michel built a benchmark harness to compare token consumption against vendor‑specific MCPs. Results showed up to 80% fewer tokens for Gong, 90% for Zendesk, 75% for Linear and a modest 16% gain on Salesforce, whose native SOQL already performs well. The test suite is open‑source on GitHub, inviting developers to probe the limits of the new layer.

Early adopters are already using the service to answer queries like “show all enterprise deals closing this month with open support tickets” or “find support tickets lacking a linked GitHub issue”. Feedback loops are open, and Airbyte hopes the community will shape the next iteration of the context layer.