How to connect Claude to BigQuery, and fix the errors you'll hit
Jump to a section
BigQuery is one of the most common data sources, so once your team adopts Claude, one of the first things you'll want is to pull that data into a conversation. It's possible, and there are two ways to do it:
- Direct connector. Claude's native BigQuery connector, set up with an OAuth client from the Google Cloud Console. Free, and best for one person.
- A connector platform like Contextflo. Connect once for the whole team, with shared context, per-person table access and an audit trail.
I tried both, so this post covers the pros and cons of each, starting with the direct connector and the errors you're likely to hit.
Attempt 1: Claude's native BigQuery connector
Claude has a native BigQuery connector. I clicked install. It asked for an OAuth Client ID and Secret.

Getting those means a detour into the Google Cloud Console:
- Set up a Google Auth Platform consent screen
- Choose internal or external audience
- Configure app name, support email, scopes
- Create an OAuth Client ID of type "Web application"
- Add the right redirect URI, but which one?
- Copy the ID and Secret back into Claude
I did all of that except the redirect URI, because nothing told me what Claude expected. So:

Google's sign-in page said:
Access blocked: This app's request is invalid
You can't sign in because this app sent an invalid request. You can try again later, or contact the developer about this issue.
Error 400: redirect_uri_mismatch
After some digging: the redirect URI Claude uses is https://claude.ai/api/mcp/auth_callback. Add that to the OAuth client, wait a few minutes for Google to propagate it, and it connects.
Still blocked after adding it? Book 20 minutes and we'll work through it with you.
Claude can now list datasets, run queries and return results. Which is where the next problem starts.
Attempt 2: Google's managed MCP endpoint
While waiting for the redirect URI to propagate I tried the other route. Google's MCP docs point at a managed server at https://bigquery.googleapis.com/mcp, so I added it as a custom MCP server in Claude.

Claude showed:
Automatic client registration isn't supported by bigquery. Edit the connector and add an OAuth Client ID.
You still need the OAuth credentials from the previous section. There's no shortcut around the consent screen.
The same docs page has one more thing worth knowing: the server isn't read-only. Google lists execute_sql as its one tool that can write, with execute_sql_readonly as the safe alternative, so connect with an identity that only has BigQuery Data Viewer and Job User. We compare this default against the other official database servers in is it safe to give Claude read-only access.
Either way, results are capped at 3,000 rows with a 3-minute timeout, per Google's MCP docs (accessed June 2026). That's fine for exploration and awkward the first time somebody asks for a full export.
Connected isn't the same as useful
The connection gives Claude your tables and columns. It doesn't tell it what they mean.
If a table is called analytics_events_v3 and a column is sts, Claude guesses. Usually plausibly, which is the problem: a wrong guess returns a number rather than an error. Two people ask "what's our revenue?" and get two different answers, and neither of them knows.
The documented fix is to add descriptions in BigQuery itself: open each table, edit the schema, describe every column.

Now do that for every column, in every table, and keep it current as columns change. It works. It also gets done for four tables and then abandoned.
Answers only stay consistent once you add that context.
When the answers trace back to you
Say you get it working for yourself. But now your ops lead wants access, your marketing manager wants campaign numbers, and your CEO wants a revenue update. And not all of them use Claude: some live in ChatGPT, one swears by Grok, a few switch between all three or reach for any other assistant that supports MCP.
Each of them goes through the same OAuth setup on their own machine, and each gets raw access to every table the credential can reach. Keeping that in order is real work. BigQuery's job history shows the SQL ran, under whichever account ran it. It doesn't show who asked, what the AI told them or which definition it used.
A contractor who needs a handful of tables means another identity, IAM grants or authorized views, and remembering to revoke it all when the contract ends.

That's also when your leaders start asking questions of their own:
- Can we trust these numbers?
- Who can see the sensitive tables?
- When an answer is wrong, how do we find out and fix it?
Getting this running, and keeping it fresh, is a full-time job even for a well-staffed data team. OpenAI's own data agent sits on six layers of context, from table usage history to code-level notes to live warehouse checks, and someone has to build and maintain every one of them.
What if there's an easier way, without giving up control?
That's the gap we built Contextflo for. You connect BigQuery once, your team asks in Claude, ChatGPT, Grok or any assistant that supports MCP, and you still decide who sees what and can trace every answer. Access rules, the audit trail and keeping context fresh come built in, instead of becoming more projects on your list. No OAuth flow, no consent screen, no redirect URI.
Want to try asking questions before connecting your own database? Explore a sample dataset and see the SQL behind the answers. No signup required. Try it on sample data
Setting it up:
- Create a read-only service account with BigQuery Data Viewer and BigQuery Job User. About two minutes.
- Paste the JSON key into Contextflo and select your project.
- Pick your datasets and tables, and Contextflo generates descriptions for each table and column. Fix what's wrong as it comes up; your team can save corrections as shared context, so it gets better with use.
- Invite your team. They connect their AI tool to Contextflo. One setup, shared.
Step-by-step with screenshots is in the BigQuery setup guide. If you need to expose views rather than tables, see restricting BigQuery access with authorized views. Planning the rollout for a whole team? The Data Lead Guide covers it end to end.
Here is Contextflo in action.
How we think about it
We built Contextflo for that person, around a few beliefs:
- Your team keeps the tool it already uses. People ask in Claude, ChatGPT, Grok or any assistant that supports MCP, not in a new dashboard they have to learn, and they all get the same answer. Contextflo sits between whichever assistant they use and BigQuery.
- Read-only, always. Contextflo only runs SELECT queries. It can't change a table, whatever someone asks.
- Every answer is traceable. Each question is logged with the person who asked it, the SQL that ran and whether it succeeded, so "where did this number come from?" takes a minute to answer, not an afternoon.
- You shouldn't have to go looking. A weekly digest sums up what your team has been asking, so you stay on top of it without reading logs.
- Context gets better with use. The descriptions start generated, and every correction someone saves becomes shared context, so a wrong answer gets fixed once for everyone.
What it won't do is the modeling. A messy revenue table becomes a well-described messy table.
If someone above you has to approve the tool, this is the section to send them.
Which one to use
Just you: use Google's connector. It's free, and once you're past the OAuth setup it does the job. You know the schema, you know what sts means, and you're the only one reading the answers.
Your team asks too: use Contextflo. The day your ops lead, a marketer or your CEO gets an answer out of BigQuery that you never saw, every number they repeat carries your name. Contextflo gives you control over who sees which tables, a record of who asked what and the SQL behind it, and context that stays fresh, without it becoming a second job.
You do the modeling, we do the rest. Connect BigQuery free, or grab 20 minutes and I'll set it up with you.
FAQ
How do I connect Claude to BigQuery? Claude has a native BigQuery connector that uses an OAuth Client ID and Secret you create in the Google Cloud Console. After you add the consent screen, create a Web application OAuth client, and paste the credentials into Claude, it can list datasets and run queries. For a whole team, Contextflo connects Claude to BigQuery once through a read-only service account instead of per-user OAuth.
Is there a BigQuery MCP server for Claude? Yes. Claude's native BigQuery connector speaks MCP, and Google also publishes a managed MCP server at https://bigquery.googleapis.com/mcp. The managed server still needs the OAuth credentials from the Google Cloud Console, since it does not support automatic client registration.
How do I fix the redirect_uri_mismatch error when connecting BigQuery to Claude? The redirect URI Claude uses is https://claude.ai/api/mcp/auth_callback. Add that exact URI to your OAuth client in the Google Cloud Console, wait a few minutes for Google to propagate it, and the BigQuery connector connects.
What is the row limit on Claude's BigQuery connector? Query results are capped at 3,000 rows with a 3-minute timeout, per Google's MCP docs. That is fine for exploration but awkward the first time someone asks for a full export.
How do I see what my team asks Claude about BigQuery? BigQuery's job history only shows that a query ran under the connecting account, not who asked or what they were told. Contextflo logs every question with the person who asked it and the SQL that ran against BigQuery, whether they asked in Claude, ChatGPT or any assistant that supports MCP, and sends the data lead a weekly digest of what the team has been asking.
What is the best BigQuery connector for a team using Claude? Contextflo. Google's connector gets one person querying. Contextflo is for when your team needs the same answer every time: it connects Claude to BigQuery once, with shared table descriptions, per-person table access, and an audit trail tying question, SQL, and person together.



