Dataki for Cloud Firestore
Ask Cloud Firestore in plain English
Query Firestore collections in plain English. Google's Firestore to BigQuery extension keeps a live copy of each collection, and Dataki answers questions about your documents from it.
Through BigQuery Google's Firestore to BigQuery extension. The guide takes about 15 minutes.
How many orders are in each status right now?
2,318 orders are shipped and on their way; 486 are still waiting for payment.
01Questions
What people ask Cloud Firestore
Each one becomes a query against the Cloud Firestore data, run read-only, with the SQL shown beside the chart so anyone who reads SQL can check it. Anything worth keeping goes on a dashboard that refreshes on its own.
- How many orders are in each status right now, and how did that change this month?
- Which users created the most documents last week?
- How many documents were deleted in each collection yesterday?
- What is the average time from order created to delivered?
02From your assistant
Ask Cloud Firestore from Claude, ChatGPT or Cursor
Dataki runs a hosted MCP server. Add it to your assistant once and it can read the Cloud Firestore data in BigQuery and run read-only queries through Dataki. The assistant gets the rows back; the credentials stay encrypted in Dataki and never reach the model.
Setup for your assistant:
From the Firestore export, how many orders moved to delivered each day this week?
- list_sources · finds BigQuery · firestore_export
- get_data_context · reads the tables and what the columns mean
- query · runs the SQL read-only, returns the rows
- render_ui · draws the chart from those rows, in the chat
https://dataki.ai/api/mcp 03Setting it up
Getting Cloud Firestore data to Dataki
Google's Firestore to BigQuery extension. The setup guide covers it step by step.
- Setup
- About 15 minutes per collection, plus a one-off import of the documents that already exist.
- Cost
- Needs the Blaze plan. The extension's function calls (2 million a month free) and BigQuery streaming inserts ($0.01 per 200 MiB) are billed to you.
- Freshness
- Seconds. Each write reaches BigQuery as it happens.
- Scope
- One collection per installation. Install it again for each collection you want to ask about.
Live dashboards
Every answer can become a live dashboard.
Ask in plain English. Dataki writes the SQL, runs it on your database and charts the answer. Keep it as a dashboard with its own link, and your team sees current numbers every time they open it.
3 views · Updated just now
One click turns the answer into a dashboard:
Always current. The query runs again whenever the dashboard opens, so Friday's meeting sees Friday's numbers.
- Share it
- Send the link to anyone. Viewers are free on every plan.
- Refine it
- The SQL is always visible. Adjust it by hand or ask for a change.
- Embed it
- Put the same dashboard in your own product with two HTML tags.
Using Dataki with Cloud Firestore
- Can I query Cloud Firestore in plain English instead of writing SQL?
- Yes. Dataki works with Cloud Firestore through its BigQuery dataset (Google's Firestore to BigQuery extension). Ask in plain English and Dataki writes the SQL against that dataset, runs it and returns a chart. The query is shown with every result, so you can check or edit it.
- How do I build a dashboard from Cloud Firestore data without an ETL pipeline?
- Follow the Cloud Firestore setup guide once and your data lands in BigQuery automatically, with no pipeline to maintain. Connect that dataset to Dataki, ask your question and save the answer as a dashboard widget. Dataki queries the dataset live, so the dashboard is always as fresh as the export.
- Can Claude, ChatGPT or Cursor query my Cloud Firestore data?
- Yes, through Dataki's hosted MCP server at https://dataki.ai/api/mcp. Connect Cloud Firestore to Dataki, add that URL as a connector in Claude or ChatGPT (or in Cursor, VS Code or Claude Code with a Dataki key), and your assistant can read the schema and run read-only queries. It works with the results, while your credentials stay encrypted in Dataki and never reach the model.