Airbyte Agents launch unified data layer for agents
Deploy Airbyte Agents to replace vendor MCPs and reduce token consumption in agent workflows.
Deploy Airbyte Agents to replace vendor MCPs and reduce token consumption in agent workflows.
Summary
Airbyte announced Airbyte Agents, a unified data layer that lets agents discover and act on information across operational systems without extensive API plumbing. The core component is a Context Store that indexes replicated data, enabling fast agentic search. Benchmarks show that Airbyte Agents use up to 80 % fewer tokens for Gong, 90 % for Zendesk, 75 % for Linear, and 16 % for Salesforce compared to vendor MCPs. The agents can read and write directly to upstream systems, and the Context Store allows structured discovery of data. The release includes a public benchmark harness on GitHub for community testing. Airbyte emphasizes the importance of indexing data ahead of time to reduce runtime token consumption. The agent example that originally took 47 steps was dramatically improved. The platform aims to accelerate agent workflows and reduce token usage.
Key changes
- Airbyte Agents provide a Context Store for agentic search.
- Benchmarks show 80 % fewer tokens for Gong, 90 % for Zendesk, 75 % for Linear, 16 % for Salesforce.
- Agents can read/write directly to upstream systems.
- Context Store indexes replicated data for fast discovery.
- Public benchmark harness available on GitHub.
- Emphasizes indexing data ahead of time.
- 47‑step agent example improved.
- Aims to accelerate agent workflows and reduce token usage.