The request lands in the inbox like clockwork. A customer success team uses your analytics dashboard for months, maybe longer. Then someone asks: Can we give this to our own users?
For most B2B software companies, the answer has traditionally involved an awkward choice—hire engineers to rebuild everything from scratch, or slap in a static embed that feels, well, slapped in. Basedash, a Y Combinator graduate building AI-powered business intelligence tools, thinks it's spotted something in between. In late July, the startup launched what it's calling a developer platform: essentially, its entire analytics engine, broken down into programmable pieces.
The bet is straightforward, if optimistic. According to the company, hundreds of teams already use Basedash internally. Now those same customers can pipe the same natural-language queries, automated SQL generation, and chart rendering directly into their own products—no rebuilding required.
Whether that proposition actually resonates at scale remains an open question.
What They're Offering
The platform works by exposing API endpoints for nearly every function Basedash performs on its own: chat interfaces, chart generation, dashboard assembly, insights, automations. A developer creates a chat session against a connected database through a POST request, streams the AI's work back as server-sent events (you can see the SQL being written and validated in real time), then pulls rendered chart images.
It's powered by OpenAI's GPT-5.6 model, released in July 2026, though the company hasn't disclosed pricing for the AI usage beyond noting it's billed separately. The streaming element is central to the pitch here—customers watch analysis unfold rather than waiting through an opaque loading state. Whether that's a meaningful UX improvement or just a nice-to-have probably depends on query complexity.
Authentication relies on API keys with a "bd_key_*" prefix, passed as Bearer tokens. For teams embedding entire applications rather than hitting individual endpoints, Basedash supports JWT-based single sign-on with short-lived tokens and origin whitelisting. The documentation is live at charts.basedash.com/api/public/, and by most accounts, reasonably thorough.
The platform builds on an embedding feature Basedash introduced in May, which let teams drop dashboards and AI chat interfaces into their applications via iframe. This goes further by making the underlying machinery programmable—you're calling functions, not just displaying a pre-built widget.
The Security Question
One of the first things people asked on Product Hunt: what happens when a customer tries to manipulate the AI into running destructive SQL?
It's a legitimate concern. Basedash's core function is translating natural language into executable database queries. In a customer-facing scenario, that means your users are essentially asking an AI to write code against your production data. Prompt injection—tricking the model into doing something unintended—isn't theoretical.
Basedash says it gates destructive actions behind an approval workflow. Human review is required before the system makes consequential changes. The details of what qualifies as "destructive" aren't spelled out in the public documentation, but the company's team members confirmed the safeguard exists.
More fundamental is the multi-tenant problem: most customer-facing analytics involve shared database tables, where Customer A shouldn't see Customer B's rows. Basedash handles this through row-level security. JWTs carry tenant identifiers, and filtering happens at both the query layer and the warehouse itself. In early June, the company published a technical breakdown of configuring RLS in Postgres, Snowflake, BigQuery, and Databricks.
The guidance emphasizes warehouse-native policies—Postgres's "FORCE ROW LEVEL SECURITY," for example—rather than relying solely on application-side filtering. It's the right architectural instinct, though implementation complexity varies wildly depending on your data warehouse setup.
Pricing and the Competitive Thicket
Basedash's current pricing lists a Startup tier at $1,000 monthly, plus AI usage fees, covering up to 25 users. Enterprise pricing is custom and includes self-hosting, deeper embedding options, and SSO through SAML or OIDC. There's some documentation drift—an older page references a $250 Basic tier—but the current marketing site appears to reflect active pricing.
The embedded analytics market is, to put it mildly, not underpopulated. GoodData, Sisense, Luzmo, Qrvey, and Explo all compete here directly. Traditional BI tools like Looker, Sigma, ThoughtSpot, Metabase, Tableau, and Power BI offer embed modes too, though they weren't designed primarily for that use case and it shows.
Basedash runs its own benchmark suite—BI Bench—claiming 92.1% accuracy with 28.6-second average response times across tasks. The test compares Basedash against ten-plus tools including Codex, Hex, Claude Code, Sigma, and Metabase. Standard vendor-benchmark caveats apply: you should read these numbers as directional, not gospel.
What's Actually Shipping
Beyond the API itself, Basedash has pushed out a few adjacent features in recent months. In early July came Actions—where the AI agent can write and execute SQL edits or trigger workflows in external tools like Stripe and HubSpot through Model Context Protocol connectors. Mid-July brought Suggestions, a proactive system that recommends questions, dashboards, and automations based on connected data and usage patterns.
The company touts SOC 2 Type II certification, encryption at rest and in transit, and integrations with Okta and Microsoft Entra ID for SCIM provisioning. Teams with strict data residency or governance requirements can opt for self-hosting or VPC deployments.
Basedash raised a $4.4 million seed round led by Matrix Partners, with Y Combinator, Form Capital, Worklife, and angels from Figma, Notion, and BloomTech participating. LinkedIn pegs the team size at two to ten employees, though only three profiles surface publicly. Founder and CEO Max Musing, a Y Combinator alum from the Summer 2020 batch, wrote the platform launch post.
The Demand Test
According to Basedash's own LinkedIn update, the pattern has played out repeatedly: companies adopt the tool for internal analytics, teams grow dependent on it, then someone inevitably asks if they can surface it to end customers. The company claims over 200 organizations as users, though that's a vendor-supplied figure without independent corroboration.
The question now is whether that informal ask—"Can we put this in front of our customers?"—translates into actual payment for the developer platform. It's one thing for a customer success team to request a feature. It's another for a company to commit engineering resources, redesign workflows, and write checks for an embedded analytics build.
The API documentation is live, and companies can start integrating immediately. Whether hundreds of existing users convert into paying embedded customers will test a straightforward thesis: that the friction of building customer-facing analytics in-house is painful enough to justify outsourcing the entire stack. In embedded BI, that's never been a foregone conclusion.
