A smarter foundation for Secoda AI

We’re at the point where AI is baked into the tools shaping modern data work. Tools like dbt, Snowflake, and others now include native AI features. They can write SQL, document a model, or spot an anomaly. All of that is helpful within the scope of each tool. But these features are limited to what each tool knows. They don’t offer a shared understanding of how your data fits together and lack the context to provide consistent, trustworthy outputs.

What sets Secoda AI apart is its ability to deliver fast, context-rich answers enabled by a foundation of key metadata, including lineage, ownership, documentation, usage, and access. We believe context is the key to making AI work for data teams, and the foundation we’re releasing today is an important step toward that vision. It brings more structure, more transparency, and better alignment with how teams actually use data to make decisions.

See the full menu of updates from this past month in our May product overview.

Secoda now uses Claude 4

We’ve recently upgraded Secoda AI to Claude Sonnet and Claude Opus 4, Anthropic’s latest models, and the improvement is immediate. Responses are now significantly faster and more consistent, with better alignment to the intent behind each question. Opus plans the response, Sonnet executes it, and together they’ve reduced latency by over 30% while improving accuracy and readability.

Introducing Secoda MCP

We’re excited to launch Secoda’s Model Context Protocol (MCP), a secure, standardized interface that lets AI tools and agents tap into your Secoda catalog wherever you work.

MCP brings your governed metadata into the AI environments you already use, like Cursor, Claude, Windsurf, and VS Code. Your existing AI tools can now access trusted context such as lineage, glossary terms, documentation, and metadata without needing to replicate logic or compromise on governance.

With MCP, external agents can:

• Search for tables, dashboards, columns, and business terms from Secoda
• Execute SQL queries against your data warehouses using catalog context
• Retrieve documentation and glossary definitions tied to your data
• Explore upstream and downstream lineage across your data assets
• Access all of this through secure, authenticated endpoints that respect workspace permissions

By enabling MCP, you extend Secoda as your single source of truth and fuel AI-assisted workflows with complete, trustworthy context.

A better experience for asking and answering questions

One of the most common challenges we’ve seen is fragmentation. A user might ask Secoda AI a question, follow up with their data team, and later check the documentation. These efforts often happen across disconnected channels.

That’s exactly the kind of disconnect we’re aiming to solve with context-aware AI. We’ve introduced a Q&A workflow that brings everything into one place:

• Users can give feedback on AI answers and flag anything that needs human review
• Each question now displays the full reasoning path, making it easier for teammates to follow the logic
• You can mention @Secoda in a reply to re-engage AI in the thread at any point
• A new AI Analytics Report helps teams track which questions are asked, how they’re answered, and what gets resolved

When a question is marked as resolved, Secoda AI learns from the answer. Over time, this improves response quality and helps the system provide more accurate guidance.

Surfacing the insights that strengthen Secoda AI

As teams ask questions and resolve issues in Secoda, the AI improves. But what truly powers accurate, trustworthy answers is the metadata behind those interactions.

To reinforce that foundation, we’ve made key signals like usage and adoption more accessible. Monitoring, Automations, Requests, and Policies now include Overview Pages that highlight:

• Weekly usage trends to show how adoption is evolving
• Time-series charts for exploring engagement over time
• Asset-level breakdowns to help teams focus where it matters

These insights give teams a clearer view of how features are being used and where they’re adding value. They also feed into the metadata graph that powers Secoda AI, helping it identify which assets are most relevant and generate smarter, more context-aware responses.

By embedding this context into the tools your team already relies on, we’re making it easier to stay aligned and continuously improve the quality of answers across the platform.

Designed for the way teams actually work

Model performance alone doesn’t determine whether AI is useful in a data organization. The real challenge is making sure AI reflects how teams operate day to day. Data lives across tools, ownership changes, documentation falls behind, etc. Without bringing those pieces together, AI struggles to offer support you can rely on.

Secoda AI is built on your metadata graph. It understands how assets connect, how terms are defined, what’s being used, and who has access. That context helps generate answers that feel familiar and useful, not disconnected from how your team actually works.

We’re building with long-term trust in mind. That means focusing on accuracy, transparency, and usability. We’re also thinking ahead. AI will continue to evolve, and we believe the most effective systems will be the ones that stay grounded in context.

Thanks for being part of what we’re building.

— Etai Mizrahi
CEO & Co-founder, Secoda