You asked for it, you got it. The Airbyte MCP server is here 🚀
Build pipelines, fix failing syncs, and use the context layer to chat with your Snowflake or BigQuery data from Claude, ChatGPT, Cursor, or any other client that supports MCP.
This was the most demanded feature in Airbyte history, and we’re excited to finally put it in your hands.
The MCP server is open source and free for every Airbyte user. The context layer is for Airbyte users on paid plans.
Read the launch blog: tinyurl.com/53k8zm64
You don't need a new database to give your agents context. Your warehouse already is one.
Here's how to turn it into a context layer with Airbyte. (continued in thread)
tinyurl.com/53k8zm64
"Treat the agent like a new team member. Least privileged by default."
Airbyte's Ian Alton on @tfir_io, on how the Airbyte MCP finds failed syncs, reads the logs, and explains the fix without ever handing the model a credential.
tfir.io/ai-agents-data-pipel…
The Airbyte MCP server and context layer are now available!
Recently, we surveyed our community (thanks to everyone who participated 🫶), and we identified there was incredible demand for an Airbyte MCP server. In fact, it was the most demanded feature in the history of Airbyte.
And it makes perfect sense! Like many of you, I don’t like switching out of my AI client to access a web UI anymore. It’s unnecessary friction.
You can now run Airbyte from Claude, ChatGPT, Cursor, or VS Code. Build a pipeline, diagnose a failed sync, or manage a connector by asking your AI agent.
Here’s what’s really cool: for paying users, the MCP server also gives you access to the Airbyte context layer. The Airbyte context layer allows you to chat directly with your Snowflake or BigQuery data warehouse. This is where your organizational context lives, and yet it is still difficult for agents to access that data for far too many teams. Not anymore.
For six years we built Airbyte for humans and dashboards. Now it can be run by the same agents that consume the data it moves.
The MCP is open source and free for every Airbyte user. The context layer is available for Airbyte cloud users.
Read the launch blog: lnkd.in/eHpjiUgd
Airbyte Enterprise Flex now supports sync logs in the UI. Search, filter, stack traces, no kubectl. Off by default; your security team should weigh in first. airbyte.com/blog/sync-logs-f…
We'll be at Data Outpost by @motherduck, Nov 4 and 5 in San Francisco 📍
Catch our developer advocate Richelle Herrli on Nov 4 for a hands-on workshop, "Turning Your Data Into AI Context, Powered by Airbyte," or find us at the booth to talk about data sovereignty... and cop an Octavia plushie!
Register here: dataoutpost.ai
ALT A digital illustration featuring a robot holding a pink flag with a "D" on a cliff. The sun is setting in a pixel-style sky. Text reads: "Airbyte is a proud sponsor of Data Outpost." Location and date info: San Francisco, November 04-05, 2026.
We had the best time at #SFTechWeek! Thank you for all of you who attended our Rooftop Refresh and made it into the party of the week 🥳
The bites were elevated. The drinks were refreshing. The people were phenomenal.
And the vibes were immaculate 💅
Our CEO @MichelTricot spoke at @TechCrunch Disrupt yesterday, and based on the absolutely PACKED 🥳 room, it's a topic many people are thinking about: what happens when OpenAI ships your roadmap?
Shared the stage with @RebeccaBellan, @YayLT, and @_RobToews
Michel shared some of his insights on how AI has been a force multiplier from Airbyte, as well as what we consider the most powerful moat for a company.
Here's a hint: it's data 🤓
Watch the full video here: piped.video/watch?v=WaM4IGvd…
We let AI agents fix our production connectors on their own.
Building it was easier than you'd think.
Trust was harder. We installed evals that prove a fix works, humans in the loop where it matters, staged rollout so one bad change stays small.
@pedroslopez broke it down at @Ai4Conferencespiped.video/watch?v=aKtpWWeB…
BigQuery is now a certified Airbyte source, our fastest one EVER!
It reads 1TB in under an hour via the Storage Read API 🤯
Typed columns, nested schemas, resumable syncs.
airbyte.com/blog/bigquery-so…
Hand your full business context to a vendor, then pay for outputs built from it. @MarioMoscat on @tfir_io calls that renting back your intelligence.
We're in the business of owning your intelligence. Check the full conversation:
tfir.io/airbyte-ai-data-sove…
700+ connectors, each on an API we don't control. That's a lot of maintenance.
@Cognition wrote up how Devin helps us keep up: ~990 issues triaged in 90 days, support resolution time roughly halved, 62% of doc updates auto-merged.
Humans still get final review. Check out the full story.
devin.ai/customers/airbyte
How we measure AI ROI at Airbyte: connectors per engineer. It started 1:1.
Hydra, our fully agentic internal system, keeps pushing that ratio higher to where we want to be.
Check out the full talk with @DevInterrupted (link below)
Your AI rollout probably doesn't have a model problem. It has a "nobody wrote this down" problem.
Our CTO Davin Chia explains this phenomenon, which we call "cognitive debt", in this piece for @sdtimes.
sdtimes.com/agentic-ai/code-…
We're throwing a rooftop party on Thursday of #SFTechWeek 🪩
Oct 8, 5:30 PM PT in SoMa. No panels, no pitches, no agenda. Just drinks, bites, music, and very good raffle prizes.
We've got 200+ signups, and we will hit capacity. RSVP now:
partiful.com/e/ADmj7eEWPS3NI…?
Connector setup rarely happens in one sitting.
Newly available draft mode for Airbyte Cloud sources & destinations lets you save, close the tab, track down the DB admin, and finish later. airbyte.com/blog/save-in-pro…
We measure AI ROI at Airbyte with one question: how many connectors can one engineer maintain?
Hydra, our fully agentic internal system, keeps pushing that number higher 📈
See the full discussion with @micheltricot on @DevInterrupted: piped.video/watch?v=iUHPcQ_7…