ChatCSV
ChatCSV turns a spreadsheet into a conversation: upload a CSV, ask questions in plain English, and get written answers plus charts on demand. Free for everyone since its acquisition by Flatfile, with a public REST API.
What is ChatCSV?
ChatCSV is a web application that lets you hold a conversation with a spreadsheet. You upload a CSV — from your computer, from a URL or straight from the clipboard — and then ask questions about it in ordinary English. The tool answers in writing and, when you ask for one, draws the chart that goes with the answer: bar chart, pie chart, histogram, scatter plot. Its own tagline sums up the intent, talking to your data the way you normally talk, with a personal data analyst on the other side.
Four capabilities are advertised on the home page. Chat to your CSVs covers the upload and the conversation itself. Quick starting questions generates a handful of obvious questions about the file you have just loaded, which removes the blank-page problem on an unfamiliar dataset. Visualise the answers produces the charts on request. Chat History keeps every conversation and lets you rename it, delete it or share it with colleagues.
The published explanation of how it works is unusually candid: ChatCSV runs Python code against the uploaded file, uses LangChain to catch and correct its own errors, and stores messages on Supabase. A public REST API mirrors the product for developers. API keys are created from the dashboard, requests carry a bearer token, the default model is gpt-4o-mini, and answers stream back as server-sent events. The API also widens file support well beyond CSV — json, xlsx, txt, html, xml, parquet, dta, sas and sav are all accepted — provided the file sits behind a publicly reachable URL. A white-label option exists and is documented through a public GitHub example repository.
The product is published by ChatCSV Technologies, Inc. and was acquired by Flatfile on 17 October 2023. It has been free for everyone ever since, with no paid tier at all. A banner at the top of the home page now presents Obvious as the successor to ChatCSV, so the tool is best judged as a working but no longer actively positioned product. A public demo loaded with the Titanic dataset can be tried without an account.
What it does
- Answer questions asked in plain English about an uploaded spreadsheet
- Draw the matching chart on request: bar chart, pie chart, histogram, scatter plot
- Suggest starting questions automatically once a file has been loaded
- Read files from a computer, a URL or the clipboard
- Summarise findings across several turns of the same conversation
- Keep, rename, delete and share past conversations
- Serve the same engine to an application through a REST API
When to use ChatCSV / When not to
A quick filter to help you decide if ChatCSV is the right fit.
When to use ChatCSV
- Analysts and business users who need answers from a CSV without writing formulas, SQL or Python
- Retail and e-commerce teams exploring sales distribution, customer behaviour and inventory data
- Finance and banking staff who want transaction trends charted from a raw export
- Marketers checking campaign performance, click-through rates and audience demographics
- Developers who want to embed a file question-answering engine through the REST API or the white-label option
When not to use ChatCSV
- Teams that need a live connection to a database, a warehouse or a SaaS tool: the only input is a file
- Organisations with procurement requirements, since no DPA, no security certification and no hosting location are published
- Users handling confidential or personal records, as submitted content feeds model training unless an opt-out is requested by email
- Buyers looking for a long-term roadmap, the product having been acquired and a successor now promoted on the home page
- People who need a mobile app, a browser extension or a non-English interface, none of which exist
How to use ChatCSV
A typical end-to-end flow, from setup to results.
- Open the public demo, already loaded with the Titanic dataset, to see the tool work without an account
- Create an account from the login page with Google, GitHub or an email address
- Upload a CSV from your computer, paste a URL pointing to the file, or paste the data from the clipboard
- Start with one of the quick starting questions generated automatically for your dataset
- Ask your own question in plain English, as you would ask a colleague
- Name the chart you want — bar chart, pie chart or another type — to get a visual answer
- Keep asking follow-up questions in the same thread, including a request to summarise the findings
- Rename, delete or share the conversation from the chat history
- For programmatic use, open the dashboard, click your account name, then API, and create an API key
- Call the chat completion endpoint with an Authorization bearer header, a messages array and a files array of publicly reachable URLs, then read the server-sent event stream
Pros & Cons
Pros
- Completely free, with no paid tier since the Flatfile acquisition
- No technical skill required: questions are asked in plain English, with no formula and no SQL
- Visual answers produced on demand, without configuring a chart
- Public demo usable immediately, with no account and no upload
- Documented REST API with Next.js and Express examples and a Postman collection
- You keep ownership of your input and receive the rights to the output
- Training on your content can be switched off on request by email
Cons
- Product at the end of its cycle: acquired by Flatfile, with a successor promoted on the home page
- No mention of the GDPR, no DPA, no Article 27 representative and no named data protection officer
- No security certification and no hosting country or region disclosed
- Content is used for model training by default, with an opt-out obtainable only by email
- Support rests on a single personal address, with no help centre and no contact page
- No pricing, about, contact or features page: those routes return a 404
- No mobile app, no browser extension and no published third-party integration
Pricing & Plans
ChatCSV is free of charge for all users and there is no paid plan: the FAQ states that the paid option was withdrawn as part of the acquisition by Flatfile. No price, currency or billing unit is published, and the site has no pricing page. There is therefore no lowest price point to report.
- Free — the only plan
- giving full access to the product at no cost
- with no paid tier since the Flatfile acquisition
- White-label option — mentioned in the FAQ and documented through a public GitHub repository
- with no commercial terms published
Data, GDPR & hosting
A consolidated view of how ChatCSV handles your data.
GDPR overview
There is no GDPR implementation to report. The word GDPR does not appear anywhere on the site, in the privacy policy or in the terms; the policy only promises to comply with "any applicable law and regulation". No Article 27 representative is designated, no data protection officer is named, no data processing agreement is published or offered, and no list of subprocessors exists — only categories of recipients such as IT providers, hosting and storage providers, advertisers and analytics platforms. Some rights are nevertheless described: withholding personal information, requesting details of what is held, having inaccurate or outdated data corrected, unsubscribing from direct marketing, and complaining to the vendor and then to a data protection authority. Both documents are dated May 2023 and governed by the law of the State of New York.
Who owns the data?
The terms are explicit on ownership. Anything you submit is your Input and you keep it; anything the tool returns is the Output, and ChatCSV assigns you all of its rights, title and interest in it, provided you comply with the terms. Both are covered by the single word Content, and you remain responsible for what you submit. Two limits follow. Output is not exclusive: the same or a similar answer may be produced for other users, and answers generated for them are not yours. And any improvement to the service derived from your Content belongs to ChatCSV alone, with no ownership share and no compensation for you.
Reuse rights
You may reuse your own Input and the Output freely, without asking permission, since the terms assign you the rights to both. The counterpart runs the other way: ChatCSV may use your Content to operate and maintain the service, to comply with the law and to enforce its own policies, and it may also use it to train, test and improve its machine learning models. That second use is the default, not an option you switch on. An opt-out exists but is manual: you email sam@chatcsv.co with your organisation ID, and the vendor warns that opting out may reduce how well the service handles your specific case. The privacy policy also allows collected data to be combined with general or research data obtained from other trusted sources.
Data retention & training
Hosting summary
No hosting jurisdiction is disclosed. Neither the privacy policy, the terms nor the documentation names a country, a region or a data centre, and the privacy policy refers only to categories of recipients such as IT service providers, data storage, hosting and server providers, without naming any of them. Two indications exist, but they are technical rather than contractual: the FAQ states that messages are stored on Supabase, and a founder tweet embedded on the home page says the service is hosted on Vercel with storage by Supabase. The domain currently resolves to an address announced by Amazon and geolocated in the United States, but that address is anycast, so it indicates a content delivery node rather than where data is at rest. In practice, anyone with a data residency requirement will find nothing on this site to rely on.
Things to keep in mind
Risks and trade-offs to weigh before adopting ChatCSV.
- Everything you upload feeds model training unless you actively email for an opt-out, which is easy to postpone and then forget
- The API requires file URLs that are publicly accessible, so a shortcut taken under deadline can expose a dataset outside the service
- No hosting country, no DPA and no certification are published, so any personal or regulated data placed here travels with no documented guarantee
- Confident-sounding answers are generated by code written on the fly; treating a chart as verified fact without checking the underlying figures is the classic way to be wrong in a meeting
- Leaning on a conversational analyst can quietly erode the habit of reading a dataset yourself, and with it the ability to notice when an answer is implausible
- Shared conversations carry the data behind them: sharing a thread with a colleague is not a neutral gesture
- The product has been acquired and a successor is promoted on the same page, so building a workflow or an integration on it is a bet on a service nobody has committed to keeping
Setup & Integrations
Technical difficulty
Very low for the web application. There is nothing to install: the demo runs without an account, and signing up takes one click through Google, GitHub or an email address. A first result arrives as soon as a file is uploaded, with starting questions offered automatically. Using the API asks slightly more: create a key from the dashboard, send an HTTP request with a bearer header, and handle a server-sent event stream on the client side. The one real constraint is that files passed to the API must be reachable through a public URL.
Deployment
Behind ChatCSV
Fundraising
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Is ChatCSV free?
Can I try it without creating an account?
Which file formats are supported?
How does ChatCSV actually work?
Is there an API?
Is my data used to train the models?
Who owns the answers ChatCSV produces?
What did the acquisition by Flatfile change?
Is there a mobile app?
Is there a minimum age?
Should you pick ChatCSV?
ChatCSV does one thing and does it without ceremony: it lets someone who is not a data analyst interrogate a tabular file in plain English and get back both a written answer and, on request, the chart that goes with it. The path in is unusually short — a public demo with no account, an upload from a file, a URL or the clipboard, and a set of starting questions generated for you when the dataset is unfamiliar. Developers get more than a marketing promise: a documented REST API, bearer authentication, streamed answers, ten accepted file formats and a white-label option with a public example repository. And it costs nothing at all, which removes the usual arithmetic from the decision.
The reservations are on the other side of the ledger, and they are not small. Nothing on the site names the GDPR; there is no data processing agreement, no Article 27 representative, no data protection officer, no security certification and no disclosed hosting country. Submitted content feeds model training by default, and the only way out is an email. Support runs through a single personal address, and the intellectual property contact is a personal Gmail account. The terms and privacy policy have not moved since May 2023.
Context matters here as much as features. ChatCSV was acquired by Flatfile in October 2023, and the home page now leads with a banner presenting Obvious as its successor. The service still works, the site promises continuity, but this is a product being maintained rather than developed. For ad-hoc exploration of a non-sensitive export, or for a quick prototype behind the API, it remains a genuinely useful and cost-free option. For personal or confidential data, or any commitment beyond the short term, the compliance gaps and the successor banner should weigh heavily.
- Choosing a selection results in a full page refresh.
- Opens in a new window.