AskYourDatabase

Converts natural language questions into executable SQL statements by encoding database schema context (table names, column definitions, relationships) into the AI model's prompt or fine-tuned weights. The system accepts user questions in English, generates SQL via Claude or GPT models, and executes the query against the connected database within a 60-second timeout window (chatbot mode) or unlimited time (desktop mode). Schema understanding is enhanced through optional 'training prompts' where users provide example natural language questions paired with their corresponding SQL queries to teach the AI about domain-specific terminology and complex join patterns.

Transforms natural language descriptions of desired dashboards into interactive, real-time visualizations containing tables, charts, and forms. Users describe what data they want to see (e.g., 'show me sales by region with a pie chart and monthly trend line'), and the system generates SQL queries, executes them, and renders the results in an embeddable dashboard component. Dashboards support multi-tenant database switching and fine-grained user-level access control, allowing different users to see filtered data based on their permissions.

Provides webhook functionality to trigger external integrations when queries are executed or results are available. Webhooks are mentioned in documentation but specific implementation details are absent — unclear what events trigger webhooks, what payload format is used, or how webhooks are configured. The system likely supports sending query results or notifications to external systems (Slack, email, custom APIs) via HTTP POST requests.

Standalone desktop application (Windows/Mac/Linux) that runs locally on user's machine with no data transmission to AskYourDatabase servers. Users connect to local or remote databases, ask natural language questions, and SQL executes on the user's machine. The desktop app includes access to Claude Haiku, Claude Sonnet, and GPT-4.1 models. No per-query timeout is documented (implied unlimited). Desktop app is licensed per-seat with single Ultimate tier ($49/month or $69.99/year) covering all features and models.

Allows removal of AskYourDatabase branding from embedded chatbots and dashboards, enabling white-label deployment. Custom branding is available at the Established tier ($329/month) and above for web chatbots. The system supports custom CSS styling and branding configuration (specific customization options not documented). Enterprise tier includes additional white-label features and custom SLA agreements.

Provides three distinct deployment architectures optimized for different security and infrastructure requirements: (1) Desktop application mode where database connections and SQL execution occur entirely on the user's local machine with no data transmission to AskYourDatabase servers, (2) Web chatbot mode where requests are sent to AskYourDatabase servers (fixed IP for firewall compatibility) which generate SQL and execute against the user's remote database, and (3) Enterprise on-premise mode where the AI model itself is deployed on the customer's infrastructure for maximum data isolation. Each mode uses the same underlying natural language-to-SQL engine but differs in where inference and execution occur.

Extends beyond SELECT queries to support INSERT, UPDATE, and DELETE operations via natural language instructions. Users can describe data modifications in English (e.g., 'update all customers in California to have status inactive'), and the system generates and executes the corresponding SQL DML statements. Access control is enforced at the user level, preventing unauthorized modifications. The system does not support DDL operations (CREATE/ALTER/DROP table structures).

Provides a JavaScript-embeddable chat widget that can be integrated into websites and web applications, allowing end-users to ask natural language questions about data without leaving the host application. The widget communicates with AskYourDatabase servers via API (Ask API, Messages API, New Chat API — specific endpoints undocumented). Additionally supports WhatsApp Business integration, enabling users to query data through WhatsApp conversations. Both channels enforce the same 60-second query timeout and question quota limits (1000 or 1500 questions/month depending on pricing tier).

System claims to improve query accuracy over time through user feedback on generated SQL queries and results. Users can provide corrections or examples of better SQL for their questions, and the system incorporates this feedback to improve future query generation. The specific mechanism (fine-tuning, prompt engineering, vector embeddings, or RAG) is not documented. The system also supports 'training prompts' where users proactively teach the AI about complex schemas and domain terminology through example question-SQL pairs.

Supports querying across 8+ different database engines (PostgreSQL, MySQL, SQL Server, Oracle, MongoDB, Snowflake, BigQuery, ClickHouse, Neon DB) through a single natural language interface. The system abstracts away database-specific SQL dialects and query optimization patterns, allowing users to ask questions in English regardless of underlying database. Each database connection is independent — no cross-database joins are supported. The system handles database-specific syntax differences (e.g., MongoDB aggregation pipelines vs SQL SELECT) internally.

Allows users to choose between different AI models for query generation: Claude 4.6 Sonnet, Claude Haiku 4.5, and GPT-4.1 (desktop/chatbot) or GPT-4 (enterprise only). All models are included in the subscription price — no per-model surcharge. Users can switch models per-query or set a default model. The system abstracts model-specific API differences (function calling conventions, context window sizes, token limits) internally.

Implements user-level permissions that restrict which data each user can access through the natural language interface. Users can be assigned different permission levels, and the system enforces these restrictions when generating and executing SQL queries. Additionally supports multi-tenant database switching, allowing a single dashboard or chatbot instance to connect to different databases based on user context (e.g., different regional databases for different users). Users can be banned/unbanned from accessing specific resources. The specific implementation (row-level security, query rewriting, view-based filtering) is not documented.

Implements hard limits on the number of natural language questions users can ask per month, enforced at the subscription tier level. Web chatbot tier includes 1000 questions/month (Scale tier at $149/month) or 1500 questions/month (Established tier at $329/month). Desktop app has unlimited questions. The system tracks question usage and prevents further queries once the monthly quota is exceeded. Overage pricing and quota reset mechanics are not documented.