Basedash Actions

Product Introduction

1. Definition: Basedash Actions is an AI-powered workflow automation and data operation module integrated into the Basedash business intelligence (BI) platform. Technically, it is an agentic AI system that executes governed write operations across databases and third-party SaaS tools via the Model Context Protocol (MCP).
2. Core Value Proposition: It exists to close the "action gap" in business intelligence by enabling AI agents to not only analyze and report on data but also to execute approved changes directly within the data source and connected operational tools. This transforms BI from a passive reporting tool into an active operational platform, automating data-driven workflows from insight to execution.

Main Features

1. AI-Powered Data Editing: This feature allows the Basedash AI agent to generate and execute SQL UPDATE, INSERT, and other data manipulation commands on connected production databases. An administrator must explicitly enable "Allow edits" on a per-data-source basis. The agent uses its understanding of the database schema and natural language prompts to write precise, context-aware SQL. For example, a prompt like "extend trial for all orgs created last week" results in a generated UPDATE statement with the correct WHERE clause, which is then presented for human approval before execution.
2. MCP Tool Actions: This feature extends the agent's capability beyond the database to any external system via the Model Context Protocol. By connecting MCP servers (e.g., for Stripe, HubSpot, Linear, Resend, or custom internal tools), the agent can perform actions like updating subscriptions, creating leads, filing issues, or sending emails. The action is performed by the agent constructing the appropriate API call payload based on the tool's MCP schema and the user's request.
3. Governed Approval Workflow: A critical security and control layer that mandates human-in-the-loop approval for consequential actions. Every data edit displays the exact SQL to be run, and every MCP tool action can be configured to show the full payload. Administrators can set permissions per tool to "Always Allow," "Needs Approval" (default), or "Blocked." This ensures no autonomous writes occur without explicit human review, maintaining auditability and control.
4. Skills for Composable Workflows: Skills are reusable, shareable instruction sets that allow users to define multi-step workflows combining data queries, edits, and external actions. A single skill can orchestrate a complex process like customer offboarding: querying for the account, updating its status in the database, canceling the subscription in Stripe via MCP, and sending a confirmation email—all as a single, governed agent operation.

Problems Solved

1. Pain Point: The "Last-Mile Problem" of Analytics. Traditional BI tools stop at visualization and reporting, creating a disconnect between insight and action. Teams waste time and context switching between dashboards, SQL clients, and various admin panels (Stripe, CRM, etc.) to manually implement changes identified by data.
2. Target Audience: Data Teams & Operations Managers (who need to automate data cleanup and bulk updates), Customer Support & Success Managers (who handle trial extensions, account fixes), Sales Operations (who provision demo environments), and Product Managers (who need to execute targeted, data-driven campaigns or fixes).
3. Use Cases: Bulk Trial Extension: Automatically identify and extend trials for users affected by a service outage. Proactive Customer Health Intervention: Identify at-risk accounts, then automatically create a task in Linear for the account manager and send a personalized check-in email. Demo Environment Seeding: One-click generation of a realistic demo org with sample data inserted across multiple related database tables. Data Correction Workflows: Fix erroneous records (e.g., incorrect user states) identified by an insight or report without writing and deploying a one-off migration script.

Unique Advantages

1. Differentiation: Unlike traditional workflow automation (Zapier, Make) that operates on simple triggers and lacks deep data context, Basedash Actions leverages a full understanding of the company's data schema and definitions. Unlike low-code internal tool builders, it uses natural language and does not require building UIs. Compared to other AI agents, its deep integration with a governed BI platform and mandatory approval layer makes it uniquely suited for secure, production data operations.
2. Key Innovation: The seamless integration of a semantic layer (reusable SQL metrics and models), a conversational AI agent, and the Model Context Protocol (MCP) within a single governed environment. This allows the agent to act with the full business context of defined metrics and connected tools, making its actions intelligent and audit-ready, not just automated.

Frequently Asked Questions (FAQ)

1. Is Basedash Actions safe for production databases? Yes, Basedash Actions is designed for governed production use. Safety is enforced through a mandatory admin-enabled "Allow edits" toggle per data source, a human-in-the-loop approval step that shows the exact SQL before execution, and comprehensive audit logs of all agent activity.
2. What is MCP and how does Basedash use it? MCP (Model Context Protocol) is an open protocol developed by Anthropic that allows AI applications to connect securely to external data sources and tools. Basedash uses MCP servers to give its AI agent read and write access to third-party services like Stripe, HubSpot, and GitHub, enabling cross-platform workflows from a single chat interface.
3. Can I build custom workflows with Basedash Actions? Absolutely. Using the "Skills" feature, you can author and save custom, multi-step workflows that chain together data queries, SQL edits, and MCP tool actions. These skills can be shared across your team, turning complex operational procedures into simple, repeatable agent commands.
4. How does Basedash Actions handle permissions and access control? Permissions are granular. Administrators can control which users or groups have access to specific data sources, MCP connectors, and individual tools within those connectors. Each tool can be set to run automatically, require approval, or be blocked entirely, providing fine-grained control over the agent's capabilities.
5. What databases and tools are supported by Basedash Actions? The data editing feature works with any SQL database supported by Basedash (e.g., PostgreSQL, MySQL, Snowflake, BigQuery). MCP actions work with any tool that has an MCP server available, including popular SaaS platforms and custom internal tools you can build servers for, offering extensive extensibility.