Wren AI

Wren AI is an open-source and commercial generative business intelligence platform designed to help people and AI agents work with business data using natural language. It combines text-to-SQL, a governed semantic layer, an AI context layer, data agents, dashboards, visualization, APIs, and integrations with modern AI clients and data platforms.

Wren AI’s current positioning centers on GenBI, or Generative Business Intelligence. Instead of requiring users to manually build every query or dashboard, Wren AI allows people and AI agents to turn business questions into governed SQL, analytical answers, charts, and dashboards. The platform is designed to operate on databases and data warehouses that organizations already use rather than requiring data to be copied into a separate proprietary warehouse.

A central part of Wren AI is its context layer. Traditional database schemas describe where information is stored, but they do not necessarily explain what business terms mean, which metrics are approved, how tables should be joined, or what organizational policies apply. Wren AI addresses this with a semantic and context layer that stores business definitions, relationships, metrics, examples, instructions, and other knowledge that AI agents can use when analyzing data.

The semantic layer is expressed through Modeling Definition Language (MDL). MDL provides structured definitions for models, columns, relationships, views, cubes, metrics, and other business concepts. These definitions can be reviewed and maintained as code, making the organization’s data context more transparent and easier to manage through development workflows.

Wren AI’s approach is intended to reduce the risk of AI systems generating plausible but incorrect SQL. Instead of allowing an AI model to interpret a raw database schema independently, Wren AI provides the agent with approved business context and then uses governed execution to produce and validate queries.

The platform supports both Agentic Mode and a more traditional Classic Mode. Agentic projects allow the system to work with capabilities such as knowledge, skills, memories, and GenBI applications. Classic projects provide a more conventional dashboards-and-spreadsheets workflow with interactive and ad-hoc analysis.

Wren AI can maintain conversational context across analytical workflows. Its newer Interactive Mode is designed for more complex, multi-step conversations in which users can ask follow-up questions, refine analysis, and continue working with the same analytical context rather than repeatedly starting new queries.

The platform can generate dashboards from natural-language prompts. Users can ask questions about business performance, marketing, finance, sales, inventory, operations, or other business areas, and Wren AI can turn those requests into analytical outputs. Its current website demonstrates workflows such as marketing-mix analysis, budget variance analysis, stockout-risk analysis, revenue analysis, and board reporting.

Wren AI also supports AI agents beyond its own interface. External AI systems such as ChatGPT, Claude, and Gemini can communicate with Wren AI through Model Context Protocol (MCP). In this architecture, the external AI agent can use Wren as a governed data sub-agent, allowing different AI clients to access the same business definitions and data policies.

This agent-native design is one of Wren AI’s major differentiators. Instead of requiring an organization to choose a single AI interface, Wren can act as a shared data and context layer underneath multiple agents. This allows organizations to use their preferred AI tools while maintaining a consistent definition of business metrics and data access policies.

Wren AI also provides APIs and SDKs for developers. Its documentation describes access through the CLI, Python packages, LangChain and LangGraph integrations, and Pydantic AI tooling. This allows developers to incorporate Wren’s governed data capabilities into custom agents, applications, scripts, and automated workflows.

The platform is warehouse-agnostic and supports more than 20 data sources. Current examples include Snowflake, BigQuery, Databricks, Amazon Redshift, PostgreSQL, SQL Server, Oracle, MySQL, ClickHouse, Athena, Trino, Starburst, and DuckDB. Wren AI’s architecture allows users to query data in place without requiring traditional ETL or migration into Wren.

Wren AI also supports data governance. Its commercial platform can apply row-level and column-level data controls at query time. This means that access policies can be applied even when a query is generated by an AI agent. Enterprise capabilities include role-based access, auditing, advanced permissions, and identity integrations.

The platform can be deployed in several ways. Wren AI Cloud provides a managed environment, while Enterprise Plus supports self-hosted deployments in a customer’s own cloud or infrastructure. Enterprise Plus also supports on-premises and air-gapped environments, making the platform suitable for organizations that cannot allow business data or AI infrastructure to leave their controlled environment.

Wren AI’s open-source edition provides developers with the context engine, CLI, semantic layer, governed text-to-SQL capabilities, MCP server, SDKs, and database connectivity. The current repository uses Apache 2.0 for the core engine, SDKs, skills, examples, and other open-source components, while documentation is separately licensed under Creative Commons.

The commercial editions build on this open foundation. Wren AI Cloud adds the hosted GenBI experience, team workspaces, dashboards, governance, and commercial API capabilities. Enterprise Plus adds the complete self-hosted platform, advanced security, SSO, SCIM, dedicated deployment, air-gapped operation, support, and enterprise controls.

Wren AI also includes a knowledge layer that can contain business instructions and examples. This helps AI agents understand organization-specific terminology and analytical conventions. The combination of structured semantic definitions and additional organizational knowledge is intended to provide a stronger foundation for AI-generated analytics than relying solely on the underlying database schema.

The platform’s AI capabilities can also be used with different models. Wren AI supports a bring-your-own-LLM approach in its open-source edition, while its cloud service provides managed model options. This gives developers and organizations flexibility in choosing the AI infrastructure that fits their requirements.

Wren AI began as a project built around Canner’s Wren Engine and Modeling Definition Language. The Wren AI project was introduced publicly in 2024, with open-source Wren AI released in April 2024 and Wren AI Cloud introduced later that year. The project evolved into its current GenBI and agent-oriented architecture through subsequent releases.

For organizations, Wren AI is particularly useful when data is distributed across multiple warehouses, databases, applications, and business systems and users need a consistent way to ask questions about that information. Its context layer can provide a common source of business definitions that can be used by employees, dashboards, APIs, and external AI agents.

Wren AI is therefore more than a conventional text-to-SQL application. Its current direction combines semantic modeling, AI context engineering, governed query execution, agentic analytics, generative dashboards, MCP, developer SDKs, and enterprise deployment. This makes it particularly relevant to organizations building AI-driven analytics systems and teams that want AI agents to work with business data without sacrificing governance or control.

Pricing

Wren AI currently offers Free, Essential Cloud, Enterprise Cloud, Enterprise Plus, and Open Source options. The Free Cloud plan costs $0 per month and includes 20 monthly free credits plus an initial 80 credits for the first 14 days. Essential Cloud costs $179 per month when billed annually and includes 13,200 annual credits. Enterprise Cloud costs $559 per month when billed annually and includes 24,000 annual credits. Enterprise Plus is self-hosted and session-based, with custom pricing.

The Cloud plans use a credit-based pricing model. Additional credits are priced at $0.10 per credit, and included credits can roll over up to two times their normal amount. Enterprise Plus instead uses concurrent-session licensing rather than monthly credits or named-user pricing.

Pricing Note

Wren AI has a permanent free Cloud plan as well as a free open-source edition. Essential Cloud and Enterprise Cloud are subscription plans with included usage credits, while Enterprise Plus is custom-priced based on concurrent sessions for self-hosted deployments. API usage uses separate API credits in the commercial Cloud environment.

Review

Wren AI is one of the more technically distinctive AI analytics platforms because it treats context and governance as core parts of AI analytics rather than simply adding a chatbot to an existing BI application.

Its strongest feature is the combination of an AI context layer with the MDL semantic layer. This gives AI agents access to business definitions, approved metrics, relationships, and organizational knowledge before they generate queries. For companies with complex data environments, this can be significantly more useful than relying on a general-purpose AI model to interpret database schemas on its own.

The open-source architecture is another major advantage. Developers can self-host the core technology, inspect the implementation, build custom integrations, use the CLI, and connect Wren to external AI agents. The Apache 2.0 licensing of the core engine and SDKs provides substantial flexibility for technical teams.

Wren’s MCP support also makes it well suited to the emerging AI-agent ecosystem. Instead of forcing users to interact exclusively with Wren’s own interface, organizations can allow tools such as ChatGPT, Claude, Gemini, or custom agents to use Wren as a governed data sub-agent. This can create a consistent data-access layer across different AI applications.

The breadth of supported data sources is another strength. Wren can work across more than 20 databases and data platforms, including major cloud warehouses and SQL databases. This makes it practical for organizations that already have a heterogeneous data infrastructure.

Governance is particularly important for enterprise deployments. Row- and column-level controls, role-based access, audit logging, SSO, SCIM, and self-hosted deployment options give organizations considerably more control than a simple consumer-oriented text-to-SQL tool. Enterprise Plus can also operate in on-premises and air-gapped environments.

The main drawback is complexity. Wren AI’s greatest strengths—semantic modeling, context management, Git workflows, agent integrations, governance, and multiple deployment models—also mean that organizations may need technical expertise to configure and maintain the platform effectively.

Pricing can also require careful monitoring because Cloud usage is credit-based and API usage has its own credit pool. Although the Free plan makes it easy to experiment, organizations with substantial usage need to understand the relationship between credits, models, questions, and API consumption.

Another limitation is that Wren is not intended to replace every traditional BI platform. Organizations seeking a simple drag-and-drop dashboard builder may find conventional BI products easier to adopt. Wren is most compelling when natural-language analytics, AI agents, semantic context, governed SQL, and programmatic data access are important requirements.

Wren AI is therefore particularly suitable for data teams, analytics engineers, developers, enterprises with multiple data sources, and organizations building AI-agent workflows over governed business data.