Evidence Core

Product Introduction

1. Definition: Evidence Core is an open-source, code-based business intelligence (BI) and data visualization framework. It falls into the technical categories of "BI as Code," "analytics engineering," and "headless BI," enabling developers and analysts to build, version-control, and deploy data products using SQL, Markdown, and a component library.
2. Core Value Proposition: It exists to replace traditional, closed-source drag-and-drop BI tools with a developer-centric, git-based workflow. Its primary value is enabling teams to build business intelligence as code, which allows for automation, collaboration with coding agents (AI), robust version control, and deployment flexibility that traditional SaaS BI platforms cannot match.

Main Features

1. Everything as Code: The entire analytics project—including dashboards, data models, access controls, and themes—is defined in human-readable files (Markdown, YAML, SQL) within a version-controlled repository. How it works: Analysts author pages/growth.md files that mix Markdown narrative with SQL queries and Evidence-specific components (like {% big_value %}). This repo-centric approach enables Git workflows like branching, pull requests, and CI/CD pipelines for analytics.
2. Analytics Agent Framework: Evidence provides a structured environment to build and deploy AI coding agents specialized for analytics. How it works: The agent is given context from the project's codebase (reports, schemas) and can be extended with custom skills (analytical playbooks) and instructions. It operates via the Model Context Protocol (MCP), allowing it to work in Claude Desktop, ChatGPT, or Slack to answer questions and generate new reports directly from the Evidence project context.
3. Polished, Interactive Data Products: Beyond static reports, Evidence supports building complex, app-like analytical experiences. Key technologies include a library of pre-built components (charts, tables, maps), custom HTML/JavaScript components, dynamic filters (date pickers, dropdowns), and features like scheduled email delivery, PDF/XLSX export, and embedded analytics via API.
4. Enterprise-Grade Deployment & Security: Designed for production use at scale. It offers page-level access control managed in access.yaml files, multi-tenancy support for customer-facing analytics, white-labeling through extensive theming, and enterprise integrations including Single Sign-On (SSO), SCIM provisioning, SOC 2 compliance, and external key management (EKMS).

Problems Solved

1. Pain Point: The "black box" nature and vendor lock-in of traditional drag-and-drop BI tools (e.g., Tableau, Power BI), which hinder automation, create governance challenges, and make collaboration with modern AI coding agents difficult.
2. Target Audience: Analytics Engineers who need to productionize data products; Data-Aware Development Teams building embedded analytics; Product Managers & Operations Leads requiring self-service reporting; and Forward-Looking Data Teams seeking to integrate AI agents (like Claude, GPT) directly into their analytical workflow.
3. Use Cases: Internal BI Platforms: Building a centralized, version-controlled source of truth for company metrics. Embedded Customer Analytics: Seamlessly integrating dashboards into a SaaS application with white-labeling and multi-tenancy. AI-Augmented Analytics: Deploying a company-specific AI agent that can answer data questions using the context of existing reports and build new ones. Governed Self-Service: Allowing business teams to create reports via Markdown/SQL in a safe, git-controlled environment.

Unique Advantages

1. Differentiation: Unlike monolithic SaaS BI platforms, Evidence Core is a developer framework. It compares to tools like Lightdash or Metabase by being fully code-native and agent-native, not just offering a "Git sync" feature. Unlike writing custom React dashboards, it provides a high-productivity, opinionated framework specifically for BI.
2. Key Innovation: Its deep integration of the "BI as Code" paradigm with the Model Context Protocol (MCP) for AI agents. This combination allows the entire analytics codebase to serve as both the production application and the context window for AI agents, creating a closed-loop system where agents can understand, query, and extend the BI platform itself.

Frequently Asked Questions (FAQ)

1. Is Evidence Core really free and open source? Yes, Evidence Core is a fully open-source framework (MIT license) that you can self-host for free. The commercial product, Evidence Studio, offers a managed cloud platform, enterprise features, and an integrated development environment.
2. How does Evidence Core work with AI and coding agents? Evidence Core structures your BI project as code, making it interpretable by AI. It supports the Model Context Protocol (MCP), allowing agents like Claude to read your project's context (schemas, reports) and execute skills to query data and generate new Markdown/SQL reports directly within the codebase.
3. What databases does Evidence Core connect to? Evidence Core supports major cloud data warehouses and databases via source plugins, including Snowflake, BigQuery, PostgreSQL, DuckDB, and more. It connects directly to these sources to run the SQL defined in your project.
4. Can I embed Evidence dashboards into my own application? Yes, Evidence provides a JavaScript SDK and API for embedding individual charts or full pages into external web applications, making it a strong solution for building customer-facing or embedded analytics products.
5. How is access control managed in a code-based BI tool? Permissions are defined declaratively in a version-controlled access.yaml file. You can specify which users, groups, or tenants have view or edit access to specific pages, enabling precise, auditable governance alongside your application code.