Today we're launching Airbyte Agents.
Itβs a platform meant to solve what we call the discovery gap.
Agents in demos are successful because they know the fields and sources they have to call to solve their task. Agents in production fail because they donβt have that knowledge assembled beforehand.
Airbyte Agents, powered by the Context Store, fixes that with one unified layer of pre-materialized context across all of your business systems.
Real results. 80% fewer tokens. 40% fewer tool calls. Get more out of your AI toolβs plan limit.
Three ways to build: MCP, SDK, or no-code Automations π‘
We're early, and we want to build with our users. Try Airbyte Agents today.
https://t.co/GcjtGBDeX7
Dallas spent three weeks in Codex with GPT-5.5. The surprise wasn't the model. It was the surface around it.
Claude is mode-first. Codex is work-first. For someone who switches between tasks dozens of times a day, this matters a lot.
The experience is what wins π‘
https://t.co/wziACpjWYK
Another great partner we've got on board.
Check out @AirbyteHQ, the data and action layer for AI agents.
Airbyte started with the open standard for data movement, powering analytics for thousands of companies.
They have launched Airbyte Agents: a context infrastructure for agents to access all your business knowledge from a single source of truth.
Excited to be sponsoring this awesome event run by @tokensandai
Show off your building skills as you work on production-ready agents. A lot of awesome companies are sponsoring alongside us including @AnthropicAI and @AWS.
π΅ $40K+ in prizes (cash, credits, giftcards)
π June 12 Β· SF
ποΈ Teams of 4 max
https://t.co/Upn4nLH46b
See you next time @Snowflake Summit!
Thanks to everyone who packed the booth to talk Airbyte Agents. And to those who caught the Airbyte bus out on the streets: yes, that was us β¨
Our happy hour was the highlight of the week. Best conversations happen off the show floor.
Next stop: @aiDotEngineer World's Fair. See you there. π
@forrester ran a Total Economic Impact study on Airbyte. The findings:
239% ROI. Payback under 6 months. $1.6M in three-year benefits π€―
Building a pipeline: 60% faster. Maintaining one: 40% faster. One team cut an ads ingestion pipeline from a month of custom code to two days.
The detail we love most: a staff engineer delivering a holiday season with zero data questions, for the first time in years!
Read it here: https://t.co/JZEg7SZTYr
DataGOL makes client data AI-ready so companies can launch agents. Step one is always the same: the data is everywhere.
Before Airbyte, that meant bash scripts, Airflow, Java, Python, and cron jobs holding it together.
With Airbyte, self-hosted on Kubernetes: 600+ connectors, built-in orchestration, one dashboard. New connectors now take hours, not days π
https://t.co/JlOksWM5fG
If you're hosting an MCP server, the auth layer is where the real work hides.
We wrote up how we built authentication for Airbyte's hosted MCP. OAuth 2.1, OIDC, FastMCP's OIDCProxy in front of Keycloak, plus the token flow gotchas nobody warns you about.
This is a must read for anyone building agentic infrastructure.
https://t.co/SOUKpsLD8d
One of our enterprise AEs vibe coded a deal tracker app!
Nai wired Gong, Calendar, Slack and his CRM through Airbyte's MCP, then built one screen showing where every deal stands.
Pre-call prep went from 15 minutes to a few π
Our MCP is making agents accessible to every function.
https://t.co/NRx6JEZSFN
We just shipped the Airbyte Agent CLI!
One standalone binary your agent runs in the shell to query and act upon your data systems. With schema discovery built in.
MCP, SDK, API, now CLI. Pick the interface that fits your agent.
https://t.co/dlUCr4Qfe6
The Airbyte Agent CLI is here β¨
One standalone binary brings the Airbyte Agents functionality to your terminal. List your workspaces, add connectors and actions like reads, searches, creates and updates. JSON in, JSON out.
Agents can even discover the schema of an operation prior to calling that operation π‘
https://t.co/U1qdfEhqHK
@BusinessInsider wrote about Airbyte and the Context Store this week π¬
The piece breaks down the five API call trap that kills agents in production, and how we're fixing it at the data layer: pre indexed, unified context ready before the agent runsπ‘
https://t.co/OqtSflbKKZ
We're kicking off Snowflake Summit this week! Hosting two fun events:
πββοΈ High Tide at the Dive with @motherduck@_hex_tech@braintrust@AirbyteHQ at the Jewish Contemporary Museum | 6/1 7-10pm
π» FDE Happy Hour with @Snowflake@modal at Kona's Bar | 6/4 3-5pm
RSVP in the comments! π