Wren AI

🚀 14,000 GitHub stars. Top 10 Trending. Same day. Today feels special. Wren AI just crossed 14,000 ⭐ on GitHub, and we’re also featured as one of the Top 10 Trending on GitHub. We’re just getting started! If you’re curious where BI is heading next, come build with us 👇 ⭐ Star → 🧠 Try → 🚀 Ship faster na2.hubs.ly/H03tDxS0

Wren AI is on GitHub Trending today. 🚀 ⭐ na2.hubs.ly/H06LW3V0 The ranking will disappear. The signal behind it won’t. Developers are building AI agents that need to do more than generate SQL. They need to understand what revenue means, which joins are valid, what they are allowed to see, and which corrections should carry into the next question. A schema can tell an agent that `rev_net` exists. It cannot tell the agent whether Finance trusts it. That is the gap we have spent years building for. Wren AI gives agents an open context layer: version-controlled metrics, relationships, instructions, memory, and access policies. The query is output. The durable asset is the business context behind it. 16,000+ GitHub stars do not mean we have solved every part of this. They do tell us we are not alone in believing BI is moving beyond text-to-SQL toward agents that can produce governed answers and deployable dashboards. Thank you to every developer, contributor, and early adopter building this with us in the open. If you are building an agent on business data, where does it break first: query generation, business context, or governance? ⭐ na2.hubs.ly/H06LW3V0 #WrenAI #GenBI #OpenSource

Most AI BI tools are a chat box. Ask a question, get one answer, start over. That is not how analytics work gets done, so we did not build that. Wren AI is an agent, not a prompt window. It operates in a sandbox and does multi-step work: it queries your warehouse, generates charts, extracts figures from PDFs, assembles dashboards, and saves the steps it takes as reusable skills. Every step is traceable. Every run is replayable. When the agent produces a number, you can open the trace and see the exact path it took to get there, then run the whole thing again on next month's data. A single-shot chat wrapper cannot do this. It has no memory of the last step and no plan for the next one. It responds, and the context evaporates. The pattern is already proven elsewhere. Claude and Cursor showed that an agent which plans, acts, and verifies beats a tool that answers once. Analytics deserves the same design. We think the chat box was a detour. The real product is an agent that does the work and shows its work. What would your team ask an analyst that a search box could never handle?

AI agents are becoming more capable every day. But for enterprise data teams, capability alone is not enough. When an AI agent answers a business question, generates SQL, creates a chart, or recommends an insight, teams need to know: Can we trace how the answer was produced? - Can we understand the reasoning behind it? - Can we evaluate whether the result is correct? - Can we improve the system when something goes wrong? That is why we are excited to introduce Thread Tracing & Evaluation in Wren AI. This release is an important step toward making Agentic GenBI production-ready for enterprise teams. With Thread Tracing, teams can inspect the full reasoning path behind an AI-generated answer, from user intent to SQL generation to final response. With Evaluation, teams can measure answer quality, compare outputs, identify regressions, and continuously improve the AI system over time. Because trust in AI does not come from a better demo. Trust comes from visibility, repeatability, and measurable improvement. As AI agents take on more analytical work, the future of BI will not only be about generating answers faster. It will be about building AI systems that can be inspected, evaluated, corrected, and trusted. This is the foundation of enterprise-ready Agentic GenBI. Read the full blog post: na2.hubs.ly/H06pfMq0

Last week we shipped a big upgrade to the GenBI app on Wren AI Cloud. It's one of the improvements our customers have asked for most. Same model, same questions, same answers, a fraction of the cost and a fraction of the wait. The biggest wins in production AI agents don't come from a smarter model. They come from sending the model less, and sending it the right less. The numbers from our internal runs for improving the GenBI Apps: - 95% fewer input tokens - 70% fewer output tokens - 90% less credit consumed - 5x faster A 5x speedup you can feel. A 90% cost cut you can see on the bill. None of it required a new model. Go check out yourself: na2.hubs.ly/H06mfG80

Our new blog post is out! This time, about our BIGGEST updates for Wren AI Open Source! We believe that within two years, most of the questions asked of your company's data won't come from a person. They'll come from agents. Claude, in your analyst's terminal, a copilot inside your product, a vendor's agent in your workflow, each one reasoning over your numbers and acting on the answer. Here's the problem nobody has solved yet: an agent doesn't know what your data means. Point it at a raw warehouse, and it guesses. It assumes `rev_net` is the number of finance trusts, invents a join, quietly includes test accounts nobody warned it about, then answers with total confidence. We call that "hallucination." The truth is, we never gave it the context to be right. A company where forty agents each invented their own definition of revenue isn't data-driven. It's forty confident, conflicting wrong answers, produced at machine speed. This is the bet we are making at Wren AI. GenBI is not a dashboard, and not a chat box. It's an open context layer your agents reason through, where what "revenue" means lives in files you own, so every agent gives the same governed answer. We just posted our new vision of Wren AI OSS: : na2.hubs.ly/H06gygG0 We built it fully open, Apache-2.0, because the layer that tells every agent what your data means is far too important to be a black box you can't inspect, fork, or own. Every agent in your company, the one your analyst runs and the one your vendor ships, is all reading the same definition of your business. No guessing. That is the world we are building toward, and it's the only thing we are building now. GenBI for AI agents. We believe it's the shape of the next decade of BI. What would your agents get wrong today if you pointed them at your warehouse? If this is the future you want to help build, three ways in: ⭐ Star Wren AI OSS: na2.hubs.ly/H06gycX0 💬 Contribute and join the community on Discord: na2.hubs.ly/H06gz7d0

The next generation of BI will not be built around dashboards. It will be built around trusted intelligence. Today, business users and AI agents both need the same thing: an accurate, governed, business-aware window into enterprise data. That is what Wren AI provides. Wren AI maps your data structure, encodes your business rules, and enforces access control — so every human and every AI agent can query from the same trusted foundation. No SQL bottlenecks. No guessing. No disconnected answers. Just plain-English questions routed to the right data, with governance built in from the start. Stop guessing your data. Start knowing it. Watch our new video and see why Wren AI is building the foundation for Agentic GenBI. 👉 Sign up now: na2.hubs.ly/H06bBDz0 #WrenAI #GenBI #AgenticBI #AIAnalytics #BusinessIntelligence #DataGovernance #EnterpriseAI

Last week at COMPUTEX, we had the opportunity to share Wren AI with people from all around the world. And what a week it was. From healthcare and hospitals, to manufacturing, retail, logistics, and many other industries, we met teams who are all asking a similar question: How can AI help us turn data into real decisions, faster and more reliably? Throughout the event, we were deeply encouraged by the conversations we had. People came with real use cases, real operational challenges, and a clear desire to bring AI into the core of how their organizations work. For us, COMPUTEX was not just an exhibition. It was a reminder that the future of AI is not only about models or infrastructure. It is about helping people across industries access trusted insights, make better decisions, and move their businesses forward. We are truly grateful to everyone who stopped by, shared feedback, asked thoughtful questions, and inspired us with their vision. To recap this wonderful week, we put together a short video capturing some of the best moments from Wren AI at COMPUTEX. Thank you, COMPUTEX. Thank you to everyone who joined us. This is just the beginning. 🚀 #WrenAI #COMPUTEX #AI #GenerativeBI #AgenticAI #BusinessIntelligence #DataAnalytics #EnterpriseAI #HealthcareAI #ManufacturingAI #RetailTech #LogisticsTech

🚀 Introducting Agentic GenBI 🔥 For two years, the industry sold you a chat box and called it AI for analytics. Every one of those demos dies at the same moment: the second question. You ask, "What was revenue last month?" It works. You ask, "Now break that down by the segments we actually care about," and it falls apart. The agent forgot the first answer, and they never knew what "revenue" meant at your company in the first place. That second question is why we rebuilt Wren AI. 🤯 The problem was never the model. A "chat with your data" wrapper breaks in four ways at once: it has no semantics or memory, it can't handle multi-step work, and governance is bolted onto the UI. None of that gets fixed by waiting for a smarter model. What's missing isn't even a semantic layer — it's a context layer. Semantics is one slice. An agent also needs to know which definitions are trusted, which joins are allowed, what your team corrected last week, and when to stop and ask. So we built the thing that does the work: an agent that reasons in steps, remembers what it learns, and stays inside your governance the whole time. We call it Agentic GenBI, and we open-sourced the core. Three years out, the dashboard is disposable. The durable asset is your context layer. The company that owns its context owns its business logic. Check the latest blog post from our CEO 👇 na2.hubs.ly/H05VWwx0

We’re heading to COMPUTEX 2026! 🚀 From June 2 to June 5, the Wren AI team will be at COMPUTEX / InnoVEX. This year, we’re excited to share what we’ve been building around: The Context Layer for AI Agents. As AI agents become part of real enterprise workflows, schema alone is no longer enough. Agents need context: trusted definitions, business logic, join relationships, reusable instructions, memory, and skills that help them reason over data reliably. At Wren AI, we’re building an open context layer that helps AI agents understand business data with governance, traceability, and reusability. We’ll also be giving two keynote sessions: 📍 June 2, 16:00–16:20 The Missing Layer: The Semantic Context Wren AI Builds for AI Agents Speaker: Jax, Wren AI 📍 June 3, 14:00–14:20 Beyond the Semantic Layer: Redefining the Context AI Needs to Understand Business Data Speaker: Paul, Wren AI You can find us at: 📍 TaiNEX 2, 4F 📍 Booth S0524 📅 June 2–5 If you’re attending COMPUTEX, come say hi. We’d love to talk about AI agents, GenBI, semantic layers, and the future of open-source AI data infrastructure. See you at COMPUTEX! #COMPUTEX2026 #InnoVEX #OpenSource #WrenAI #AIAgents #GenBI #SemanticLayer

We just launched the new Wren AI with tons of exciting capabilities, now built around "The Agentic GenBI". 🔥 This is not just “chat with your data.” Wren AI agents can now reason through multi-step analytical tasks in a sandboxed environment: query data, create charts, extract from PDFs, build dashboards, save reusable skills, and make every step traceable and replayable. A few things we are especially excited about: ✅ Agentic Design   Sandboxed, multi-step reasoning for real analytical workflows — not just one-shot Q&A. ✅ Skills & Memory   Reusable workflows and context that compound over time. ✅ GenBI Apps   Vibe-code dashboards from a single prompt. ✅ Git-native MDL   Semantic models, instructions, skills, and memory live as files that agents can read, write, version, branch, review, and roll back. ✅ Unified Data Policy   RLS, CLS, role-based access, activity logs, and governance enforced at query time — across UI, API, and MCP clients. ✅ Agent-agnostic Architecture   Works across humans, apps, and agents — governed at execution. The future of BI will not be dashboards alone. It will be agentic, governed, and deeply contextual. That is what we are building with Wren AI. Check it out, try it yourself: getwren.ai 🙌

Big news: Wren AI now supports MCP 🚀 This means your AI agents can now connect directly to Wren AI and work with your business data from the tools you already use — including Claude Code, Claude Desktop, and ChatGPT. With WrenAI MCP, your agent can: ✅ Ask business questions in natural language ✅ Generate SQL ✅ Run SQL against your data ✅ Build charts ✅ Summarize results ✅ Turn raw data into business insights All through your existing Wren AI project. This is a big step toward our vision for agentic Business Intelligence. For the past decade, BI has mostly meant opening dashboards, writing SQL, waiting for reports, or asking data teams for help. But the next generation of BI will look very different. You won’t always “go to BI.” BI will come to where you work. The future of BI is agentic. And it is already here. docs: docs.getwren.ai/cp/guide/integ… #WrenAI #MCP #GenBI #BusinessIntelligence #AIAgents #DataAI #ClaudeCode #ChatGPT

🚀 Wren AI just passed 15,000 GitHub stars! What a milestone. When we started Wren AI, we believed the future of BI would move beyond static dashboards — toward conversations, context, and AI agents that truly understand business data. Today, 15k stars feels like a strong signal that the market is moving in that direction. And here’s a small spoiler: Wren AI 2.0 is on the way. We’re preparing a major update across both Wren AI Cloud and Wren AI Open Source. More powerful. More semantic. More agent-ready. Built for the next generation of BI. To everyone who starred, tried, contributed, shared feedback, or introduced Wren AI to your team: Thank you. This is just the beginning. Stay tuned. Big things are coming. 👀 ⭐ GitHub: na2.hubs.ly/H058Dk00 #OpenSource #GenBI #BusinessIntelligence #AI #DataAnalytics #TextToSQL #WrenAI

🚀 Big news: Wren AI × Admazes are joining forces across Greater China, Hong Kong, Singapore! We’re excited to announce a strategic partnership between Wren AI and Admazes, expanding our partner network across Greater China, Hong Kong, and Singapore. Why this matters (and why now 👀) AI in analytics has promised a lot, but too often stops at demos. This partnership is about shipping real outcomes. Admazes chose Wren AI for what enterprises actually care about: 1. Enterprise-grade privacy & auditability 2. Governed, trustworthy GenBI 3. Measurable business lift not just dashboards By combining Wren AI’s Generative BI (GenBI) with Admazes’ proven MarTech playbooks, we’re enabling teams to move from insight to action fast: All of it: scalable, governed, and production-ready. 🙌 The bigger picture This partnership marks an important step in Wren AI’s mission to make GenBI the default analytics layer for modern enterprises—now with deeper local reach and execution power across Asia. 👉 Insight → Decision → Action. No friction. No black boxes. See the full announcement here: na2.hubs.ly/H04SP900

Most data questions die in Slack. 💀 A revenue ops lead asks "what's our Q1 pipeline by segment?" and now a data engineer has to stop what they're doing, open a BI tool, write a query, screenshot the result, and paste it back into the thread. 20 minutes for a 5-second answer. Multiply that across a 200-person GTM team and you've got a full-time analyst just playing messenger. We've seen this loop at SaaS companies, fintechs, e-commerce teams. The question is always simple. The answer always exists in the database. But the friction of switching contexts kills it. The best analytics tool is the one your team is already in. For most companies, that's Slack. Now you can ask data questions directly in Slack and get answers in the same thread. SQL generated, query executed, result returned. No tab-switching. No waiting on someone's queue. See it in action 👇 What makes this different from a chatbot bolted onto a database: 1/ Semantic layer underneath. The AI isn't guessing your schema. It understands your business logic, metrics definitions, and table relationships. Whether your data sits in Snowflake, PostgreSQL, or Databricks (any databases and data warehouses). 2/ Accuracy you can tune. Your data engineers can teach the AI correct SQL and validate results before rolling it out to the team. That's the difference between a demo and a real deployment. Now you can ask data questions directly in Slack and get answers in the same thread. SQL generated, query executed, result returned. No tab-switching. No waiting in someone's queue. No new tool to adopt. Your team already lives in Slack. Meet them there. The hardest part of data democratization was never the SQL. It was the context switch. What's the most common data question your team asks in Slack that still gets routed to a human? 💬 Try it free: na2.hubs.ly/H04N66J0

We're excited to announce our partnership with Zetta. Zetta turns fragmented operations into unified, intelligent systems — connecting siloed data, teams, and workflows into a single operating environment using data integration, workflow automation, and applied AI. They operate across North America, Latin America, the Middle East, and South Asia. Wren AI brings GenBI (Generative Business Intelligence) the semantic layer that makes unified data talkable. Together, we're making it possible for teams across four continents to ask questions in plain language and get trusted, context-aware answers from their operational data. No dashboards to build. No SQL to write. Before: Integrate data → build dashboards → repeat per region. Now: Define the semantics once → anyone asks, anywhere. This partnership brings GenBI to real operations, in new geographies, at scale. We can't wait to show you what's next.

Introducing "Embedded Threads" 🎉 This is one of the top requests we received from our customers, and we delivered! Dashboards were never the final interface for data. It’s about embedding governed, AI-native decision-making into the flow of work. That’s why we built Embedded Threads in Wren AI. Now teams can embed conversational analytics directly inside their own product, customer portal, or internal app. Instead of sending users to a separate BI tool, you can let them: - ask business questions in natural language - generate charts instantly - explore metrics in context This is bigger than embedding a chat window. It’s about embedding governed, AI-native decision making into the flow of work. For SaaS companies, this means shipping analytics experiences faster. For enterprises, it means bringing secure, identity-aware data conversations to every team. We believe the future of BI won’t live in a separate tab. It will be embedded, conversational, and context-aware. That’s the direction we’re building with Wren AI. Sign in and check out! 👉 na2.hubs.ly/H04wGbx0 docs: na2.hubs.ly/H04wG600

AI agents are only as powerful as the data they can actually understand. Today we made that part ridiculously easy. 🤪 With ONE command, you can install the Wren Engine skill in Claude Code and start talking to 15+ databases in minutes. And it's open-sourced and 100% FREE! 🤯 No pipelines. No glue code. No custom connectors. Just ONE command install → connect → ask. Suddenly, your AI agent can query across: Snowflake Databricks BigQuery Redshift PostgreSQL MySQL SQL Server Oracle DuckDB Spark Trino S3 / GCS / MinIO Local files …and more. All through a unified context layer built for AI agents. Under the hood: • Open-source • MCP-native • Built in Rust • Powered by Apache DataFusion • Designed for agent-first data access This is the open context engine for AI agents. If agents are the new interface, Wren Engine is the bridge to real data. ⭐ If this excites you, help us grow the project. Star the repo: na2.hubs.ly/H04sGlp0 Get started with ONE COMMAND! na2.hubs.ly/H04sBgJ0 Let’s build the data infrastructure for the AI agent era.

💡 New Updates! Most AI data tools are a black box. You ask, you get an answer, and you move on.But what if your AI could actually learn from your feedback? With Thread Tracing in Wren AI, every AI interaction becomes traceable and improvable: 1️⃣ Teach the AI what went wrong Hit 👍👎 on any answer or chart → tag the failure reason → link it to the specific knowledge rule the AI missed. Next time? It gets it right. 2️⃣ Track Answer and Chart quality separately The data can be correct but the visualization wrong. Thread Tracing catches both — each with its own feedback and improvement path. 3️⃣ Ｔeam-level observability One dashboard for your entire team: positive/negative rates, failure breakdown by category, and filter by reporter to see who's flagging what. This is Generative BI with accountability: not just answers, but traceable, improvable, governed AI analytics. 👉 Try Wren AI Cloud today：na2.hubs.ly/H04blRl0