Basedash AI Kit

Product Introduction

1. Definition: The Basedash AI Kit is a developer platform and API suite that programmatically exposes the core functionalities of the Basedash AI-native business intelligence (BI) platform. It is a technical infrastructure layer that allows developers to integrate a fully managed, conversational AI data analyst, automated insights, dashboards, and reporting workflows directly into their own applications or internal systems.
2. Core Value Proposition: It exists to eliminate the immense development burden of building secure, multi-tenant, AI-powered analytics from scratch. The kit provides a production-ready, API-first AI data analyst engine, enabling product teams to ship customer-facing analytics with a custom UI and data teams to extend their internal BI into custom workflows, all powered by the #1 ranked AI agent on BI Bench.

Main Features

1. Programmable AI Data Analyst API: This is the core feature, exposing Basedash's conversational AI analyst as a RESTful API endpoint. Developers can POST /chats to initiate a session, send natural language questions (e.g., "Why did churn increase last quarter?"), and stream back structured, validated responses. The API streams Server-Sent Events (SSE) detailing the agent's real-time progress—schema exploration, SQL generation, verification, and chart building—before delivering a final, grounded answer. It uses GPT-5.6 for reasoning, with strict server-side validation against the connected data source's actual schema.
2. Multi-Tenant, Secure Chart & Data Rendering: Every analysis generated via the API includes structured chart objects with dedicated image endpoints. This allows for native rendering within custom React/Vue components or simple image embedding in emails/PDFs. Critically, every API request is automatically scoped by row-level security (RLS) rules defined in Basedash, ensuring complete data isolation between customers or user groups without any custom frontend logic.
3. Full Platform API Coverage: Beyond chat, the API provides comprehensive control over the entire Basedash ecosystem. This includes managing dashboards and their constituent tabs, scheduling and triggering automations, defining metrics and skills, administering data source connections, and auditing AI usage logs. This turns Basedash from a standalone app into a headless BI infrastructure that can be managed as code.
4. Basedash MCP Server Integration: For AI-agent-to-AI-agent interaction, the platform offers a Model Context Protocol (MCP) server. This exposes the Basedash analyst as a tool for AI coding assistants like Claude Code, Cursor, or custom-built agents, allowing them to query live databases through the same verified, secure pipeline. This complements the HTTP API for human-in-the-loop applications.

Problems Solved

1. Pain Point: The "Quarters-Long Analytics Build." For product-led SaaS companies, building secure, scalable, customer-facing analytics dashboards with modern AI features is a massive engineering project involving query layers, charting libraries, caching, permissions, and AI agent reliability.
2. Target Audience: Product Engineering Teams at B2B SaaS companies needing embedded analytics; Data Platform & Internal Tools Teams seeking to integrate BI into existing workflows (e.g., Slack, internal wikis, support tickets); AI/ML Engineers building agents that require reliable, governed data access.
3. Use Cases: White-Labeled Customer Analytics: Embedding a fully custom, AI-chat-enabled analytics suite inside a SaaS product. Internal Workflow Automation: Automatically generating a performance report when a new deal is marked "closed-won" in Salesforce and posting it to a Slack channel. Compliance & Cost Monitoring: Pulling AI query audit logs and usage metrics into a company's central data warehouse for analysis.

Unique Advantages

1. Differentiation: Unlike traditional embedded analytics SDKs (e.g., from Looker or Metabase) that primarily embed pre-built dashboards, the Basedash AI Kit exposes an interactive, conversational AI engine. Unlike using raw LLM APIs (OpenAI, Anthropic), it provides a fully managed layer with schema-aware SQL generation, execution, validation, and chart building, eliminating hallucination and security risks.
2. Key Innovation: The "Ask. Stream. Render." loop, powered by a #1 BI Bench-ranked AI agent. The combination of a top-tier accuracy score, real-time progress streaming, server-side RLS enforcement, and a comprehensive platform API creates a unique offering. It moves beyond static dashboard embedding to provide a programmable, conversational data analysis core.

Frequently Asked Questions (FAQ)

1. How does Basedash AI Kit handle data security and isolation for multi-tenant applications? The kit enforces security at the API layer using Basedash's row-level security (RLS) system. When you make an API call with a user or customer context, every generated SQL query is automatically wrapped with RLS predicates, ensuring data is scoped correctly on the database server itself. This is more secure than relying on application-layer filtering.
2. Can I use the Basedash AI Kit to build a completely custom UI, or am I limited to iframes? Yes, the primary purpose of the AI Kit is to enable fully custom UIs. You use the API to send questions and receive structured data and chart definitions, which you then render using your own frontend components (like Recharts, Victory, or Chart.js). The iframe-based embedding is a separate, simpler product.
3. What databases and data warehouses does the Basedash AI analyst support via the API? The API supports all data sources connectable to the Basedash platform, including PostgreSQL, MySQL, Snowflake, BigQuery, Redshift, and ClickHouse. The AI agent's performance is consistent across these because it generates and validates dialect-specific SQL.
4. How is the Basedash AI Kit priced for API usage? Pricing is based on your Basedash workspace plan, which typically includes a pool of AI analyst queries ("AI credits"). API consumption draws from this same credit pool. You should consult Basedash's current pricing page for specific details on credit allowances and overage costs for high-volume API implementations.
5. What happens if the AI analyst generates an incorrect SQL query? The system includes a pre-execution verification step. Before running, the generated SQL is analyzed for potential errors, performance issues, and security compliance. Queries deemed problematic are flagged in the stream, and the agent is prompted to correct them. This validation layer, combined with execution against the real schema, drastically reduces error rates compared to raw LLM SQL generation.