Free trial
The homepage advertises a free start with no credit card required and a sample dataset to try before connecting anything. No time limit or feature restrictions are stated.
Learn moreVerbaGPT
Ask your database questions in plain English and get Python-powered analysis back
Intermediate
What is VerbaGPT?
VerbaGPT turns plain English questions into Python analysis, charts and models against your databases and files. It combines a cloud analyst (Virgo) with a local Claude Code agent (Taurus), plus Data Notes and shared prompt libraries so the model understands your schema and business logic.
VerbaGPT is a natural language analytics layer that sits on top of your databases and files. Rather than generating SQL alone, it writes executable Python: pandas for transformation, scikit-learn for modeling, matplotlib for charts. You connect PostgreSQL, MySQL, SQL Server, Snowflake, or upload CSV, Excel, JSON and PDF files (Google Drive import is also offered), then ask a question like "what were the top 10 products last quarter" and receive tables, charts and statistical output rather than a query string you have to run yourself. The product ships in two modes. Virgo is the browser-based cloud analyst: it runs at app.verbagpt.com, uses embeddings to pull only the relevant schema columns before writing code, and executes that code in a sandboxed process. It supports several frontier models from modern providers, so one model can handle schema reasoning while another writes and reviews code. Taurus is a local agent powered by Claude Code that plans multi-step investigations, runs a Python REPL against your own files and databases, corrects its own errors, and keeps the underlying data on your machine. A Taurus connector also lets the web app reach databases on your laptop or LAN without exposing them to the internet. The context problem gets its own layer. Data Notes act as a living data dictionary at database, table, column, value and sample-question level, so the model knows that status 2 means active or that customer_id joins to the clients table. Prompt Libraries capture prompts that already worked and share them org-wide, with semantic footnotes surfacing relevant examples. The workspace also learns from corrections and thumbs-up ratings, drafting Data Note improvements for you to approve, always scoped and reversible. It fits best with data teams supporting non-technical colleagues, and with consultancies that publish a database to clients through a link covered by the firm's own credit pool. It is heavier than a simple text-to-SQL chatbot and expects someone to curate context before answers get reliable.
How do you use VerbaGPT?
1Create an account and open the web app
Sign up with no credit card required and land in the Virgo workspace at app.verbagpt.com. You can start with the sample data to see how answers come back before touching your own sources.
2Connect a datasource or upload files
Point Virgo at PostgreSQL, MySQL, SQL Server or Snowflake, or upload CSV, Excel, JSON or PDF files and import from Google Drive. Combine files and a live database in the same conversation if the question needs both.3Write Data Notes for the tables you use
Describe table relationships, column meanings and coded values in plain English at database, table or column level. This is the single biggest lever on answer accuracy, so do it before inviting colleagues.4Ask a question and read the output
Type a specific question such as a monthly sales time series by region. Virgo selects relevant schema columns, writes Python with embedded SQL, executes it in a sandbox and returns charts and tables in the chat.5Correct the answer and save the fix
When a result is wrong, correct it in the conversation and approve the reusable rule the workspace offers. Thumbs-up on good results teaches it to reuse that approach and cite it under Sources.6Optional: add Taurus for local work
Install the Taurus app on your machine for local files and LAN databases, or to run private inference through Ollama or an OpenAI-compatible engine. Review the security overview first if data residency matters.
Pros and cons
Pros
- Generates full Python (pandas, scikit-learn, matplotlib) rather than only SQL, which opens up modeling and custom chartsAI
- Data Notes give the model schema and business-logic context, reducing plausible but wrong queriesAI
- Taurus keeps files and databases local, sending only the question and printed results to the modelAI
- Client sharing lets consultancies publish a dataset to outside users without per-seat fees on their endAI
- Solid governance story: isolated code execution, AES-256 at rest, role-based access and audit loggingAI
Cons
- Value depends on setup effort: Data Notes must be written and maintained before answers become trustworthyAI
- Taurus is tied to Claude Code, so local agent workflows inherit that dependency despite the model-agnostic plansAI
- The two-mode split (cloud vs local, with a connector app for LAN databases) adds real configuration overhead for new teamsAI
- No published pricing figures on the homepage, making budget planning hard before signing upAI
- Learning and prompt-sharing features work best at team scale and offer little to a single occasional userAI
How much does VerbaGPT cost?
Integrations
Databases: PostgreSQL, MySQL, SQL Server and Snowflake, including Windows sign-in for SQL Server. Files: CSV, Excel, JSON, PDF and Word, plus Google Drive import. Model providers include frontier cloud models, Ollama, LM Studio and any OpenAI-compatible engine through Private Inference.
Features
Core features include text-to-Python analysis with statistical modeling, machine learning and matplotlib visuals; Data Notes as a contextual data dictionary; Prompt Libraries with semantic footnotes; multi-model orchestration in Virgo; the Taurus local agent with multi-step reasoning; private inference via Ollama or LM Studio; client sharing with credit caps; and a workspace memory that learns from corrections with review and reversal.
Frequently asked questions about VerbaGPT
Does VerbaGPT offer a free trial?
The homepage advertises a free start with no credit card required and a sample dataset to try before connecting anything. No time limit or feature restrictions are stated.How do you use VerbaGPT?
The walkthrough on this page covers 6 steps: 1. Create an account and open the web app 2. Connect a datasource or upload files 3. Write Data Notes for the tables you use 4. Ask a question and read the output 5. Correct the answer and save the fix 6. Optional: add Taurus for local work.What platforms does VerbaGPT support?
VerbaGPT is available on Web App, MacOS, Windows and Linux.What does VerbaGPT integrate with?
Databases: PostgreSQL, MySQL, SQL Server and Snowflake, including Windows sign-in for SQL Server. Files: CSV, Excel, JSON, PDF and Word, plus Google Drive import. Model providers include frontier cloud models, Ollama, LM Studio and any OpenAI-compatible engine through Private Inference.What are the limitations of VerbaGPT?
Value depends on setup effort: Data Notes must be written and maintained before answers become trustworthy. Taurus is tied to Claude Code, so local agent workflows inherit that dependency despite the model-agnostic plans. The two-mode split (cloud vs local, with a connector app for LAN databases) adds real configuration overhead for new teams.
Status
StatusActive
Views0
Outbound clicks0
Added10/21/2026
Platforms
Web AppMacOSWindowsLinux
Pricing
FreemiumFree Trial
