Ed has been the AI assistant in the Weld editor for a while. It was useful, but it only knew what you showed it. You picked the tables, it got their column names, and that was it. Ask about anything else and it had to guess.

So we've made Ed an agent.

Now you just ask. Ed finds the tables itself, reads the columns, writes the SQL and checks it against your warehouse before you see it. If the check fails, it reads the error and fixes it. Every step shows up in the chat, so you can see what it looked at.

In short, Ed finds out the query is broken before you do.

How Ed answers a question. You ask in plain English, with no tables picked first. Ed, Weld's AI agent, works through the Weld MCP server's tools: it finds your tables, reads their columns, writes the SQL and, when it needs to, checks your data with read-only queries. Ed sends the SQL to your warehouse, which checks it without running it. If it is not valid, the error goes back to Ed, which fixes the SQL and checks again. If it is valid, you see the proposal, and Apply writes it to your open tab.

What changed

BeforeNow
You picked the tables before you askedEd searches your synced tables and models itself
Its view of your model froze when the chat startedIt sees the current SQL in the editor on every message
SQL arrived unchecked, and you found out by running itEvery proposal is checked against your warehouse first
Applying a fix meant copy, click, select all, pasteApply writes it to the open tab as one undo step
You couldn't see what it had doneEvery lookup and check shows as a step in the chat
Fix with Ed sent the bare error, and was hidden for "Reference not found"It sends the error, and the fix is checked like any other proposal
Chats were gone when you closed the panelChats are saved and tied to the tab they were about
Prompts left the EUPrompts are processed in the EU

The screenshots below follow one request from start to finish: link our Intercom contacts to our HubSpot contacts, so we can see each customer across both systems.

It finds the data itself

Ask Ed to link two of your connected sources and it lists the tables on each connection, searches for the ones that match, and reads their columns before it writes anything.

The Ed sidebar. Under the question 'I want to build a SQL model that links our Intercom data to our HubSpot data so I can see each customer across both systems', Ed shows the steps it took: listed tables from the intercom and hubspot connections, searched tables for 'intercom' and 'hubspot', and read the columns of raw.intercom_test.contact, raw.hubspot.contact, raw.intercom_test.company and raw.hubspot.account_details.

So you can skip the table picker. Describe what you want and Ed goes looking.

It checks its own SQL

Before a query gets to you, Ed sends it to your warehouse to check. Not to run it, just to ask whether it would run. If the answer is no, Ed reads the error, fixes the query and asks again.

What you get is marked Valid, with a short summary of what it does on top.

An Ed proposal card marked Valid, with a summary reading 'Creates a deduplicated, email-based customer link between Intercom and HubSpot contacts, retaining unmatched records from either system', the SQL below it, and Apply, Run, Show changes and Copy buttons. Ed's answer below explains what the model does and notes that it confirmed the SQL passes BigQuery validation.

Four things you can do with a proposal:

  • Apply puts it into the open tab as one undo step, so a single Cmd+Z (Ctrl+Z on Windows) brings back exactly what was there.
  • Run opens it in the preview, like running anything else in the editor.
  • Show changes gives you an inline diff against what's in the editor now.
  • Copy gives you the SQL as text.

Apply fixes the most annoying part. Every fix used to be copy, click, select all, paste, and hope the new version was better. Now it's one click, and Cmd+Z if it wasn't.

It sees what you're working on

The old Ed took a snapshot of your model when the chat began and never looked again. After the first message it stopped sending your SQL altogether, so you could rewrite half the query and still be talking to something that knew the version from ten minutes ago.

Now the editor's current SQL goes with every message, along with whatever you've selected. Ask it to explain or fix the selected lines and it knows which lines you mean.

It can look at the data

Some questions you can't answer from column names. What values does status actually hold? Will this join give me duplicates?

Ed can now look. It runs a read-only query, reads a small sample, and writes SQL for the data you actually have.

Whether Ed can read your data is one workspace setting, and it covers anything you connect through the Weld MCP server too. The FAQ covers who decides and what Ed can see either way.

Fix with Ed sends the error straight over

When a query fails in the preview, there's a Fix with Ed button. It hands Ed the error, and Ed checks the fix the same way it checks everything else.

Start from an empty editor

You don't need an open model to use Ed anymore. A new tab asks what you want to query, and your answer opens a draft with Ed already working on it.

The model editor's empty state, with the heading 'What would you like to query?', a prompt box reading 'Ask Ed to write it, e.g. weekly revenue by channel for the last 90 days' with an Ask Ed button, and a Blank SQL model button below it.

Each of your synced sources also has an Explore with Ed button, which points Ed at that source and lets it suggest where to start.

Your chats are saved

You can reopen a chat later instead of explaining the same model again. Each chat belongs to the editor tab it was about. Open a different tab and a proposal offers to take you back to the right one, so it doesn't land somewhere it doesn't fit. The bar above the input lets you pick up an earlier chat about the tab you're on, or start a new one.

Where this comes from

Under the hood, Ed uses the same tools as the Weld MCP server, the one you connect Claude, Cursor or Copilot to when you want to work with Weld from outside the app. Searching tables, reading schemas and running read-only queries work the same in both places. So Ed inside Weld and your own agent outside it see your workspace the same way.

Ed isn't tied to a single model, and whichever one answers, your prompts are processed in the EU.

If you haven't set up the MCP server yet, here's how.

Availability

Ed is available to anyone with permission to run queries. You'll find it in the model editor sidebar, on Fix with Ed after a failed run, and on the empty editor tab.

If you're an admin, there's one thing to decide: whether Ed can read the values in your data. It's under Settings → AI. That's it.

The Ed documentation has the full detail. You can also read the October release notes, or open Ed in your workspace.

FAQ