# Your Data Team's Claude Code Moment with Collate 2.0 AI Native Experience

Jun 10, 2026

![](data:image/svg+xml,%3csvg%20xmlns=%27http://www.w3.org/2000/svg%27%20version=%271.1%27%20width=%2748%27%20height=%2748%27/%3e)![](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1713287092723/24a64e01-4870-488f-9ddd-56aeb1cc4a54.jpeg)

James Nguyen

![Your Data Team's Claude Code Moment with Collate 2.0 AI Native Experience](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)

![Your Data Team's Claude Code Moment with Collate 2.0 AI Native Experience](https://cdn.hashnode.com/res/hashnode/image/upload/v1780604884183/7fe4b0cd-39ef-431d-a4ae-1fc34e33d4d8.png)

AI for code worked because code already lived inside a designed context system. Files had structure, dependencies were explicit, version history was captured. The software developer was the first persona to get a real AI workflow.

The data team should be next. Instead, analysts hunt across spreadsheets to confirm a metric. Stewards re-explain the same definition every quarter. Data engineers triage failed pipelines through tools that do not talk to each other. The tooling is siloed, and the AI assistant bolted onto each tool is just as fragmented in its understanding.

Collate 2.0 reimagines a new AI Native experience for data teams. Chat replaces navigation. A personalized landing page automatically shapes itself to each persona. Recurring work becomes a natural-language LLM-driven automation. Human teams and AI reason together in collaborative chats, over governed context. The corrections a steward makes in conversation become organizational memory the next agent inherits. Plugins and Skills, composed in AI Studio, assemble into the custom agents for your team. The whole experience runs on an open-source context layer from the OpenMetadata project, bringing the context, semantics, and organizational memory of your data landscape into one place.

This is the data team's Claude Code moment, as Collate brings AI for Data teams.

## Why the data team got bolted-on AI, not native AI

Every data catalog vendor in the past three years added a chat box. However, a simple chat box on top of a static UI does not change the work. An analyst still has to know which dashboard to open, a steward still has to click through a pending-approvals list, an engineer still has to triage a failed pipeline through three tools and a Slack thread.

The deeper problem is structural. The platforms that data teams run on were not designed for an agent to act inside them. Information lives in disconnected views, and workflows depend on someone remembering where to click next. The model cannot find what it cannot see, and cannot improve what it cannot remember.

When OpenAI built their internal data agent on top of OpenMetadata, the system covered 4,000+ employees, 600+ petabytes across 70,000 datasets, and six layers of context the agent reads. The lesson they took away was clear: the model is not the bottleneck. The context system around it is. The same insight drove the foundation for Collate 2.0, so that AI for the data team can work the same way AI for code does. The platform has to be designed for the model to act inside it, for both its human and machine data teams.

## Introducing the AI Native Experience

![Connect, Reason, Remember — the three primitives of the AI Native Experience](https://cdn.hashnode.com/res/hashnode/image/upload/v1780604891162/70c01370-997e-4b47-8c84-2c4efb8710bf.png)

At the core of Collate 2.0 is a redesigned experience built to give both humans and AI agents the same context layer to connect, reason, and remember across your data.

**Connect** is the open knowledge graph that ties every source, asset, lineage edge, glossary term, and policy into one model. 130+ native connectors map every cloud and platform, and reverse metadata writes governance decisions back into the source systems they came from.

**Reason** is the semantic layer that grounds every data asset in business meaning. Glossary terms, ontology, metrics, and data products give the agent the same understanding a senior analyst has, across relationships, not just labels.

**Remember** is the organizational memory layer that captures corrections, definitions, and preferences once and makes them available across every user, agent, thread, and asset that follows.

Together, the three primitives sit on the open context layer, developed in open source and built on open industry standards. The result: the data team's day starts with what matters, the recurring work runs itself, and every conversation compounds the team's knowledge instead of disappearing into a Slack thread. Let's work through the capabilities of this new AI Native experience.

## Personalized landing pages, actionable for every persona

![Personalized landing page tailored to each data persona](https://cdn.hashnode.com/res/hashnode/image/upload/v1780606675656/96c195a7-542c-4825-a899-ce996f8b8493.png)

Open Collate 2.0 and you land on a view shaped to your role. The data engineer sees a Platform Health Score, what is failing, and what to fix first. The governance lead sees pending approvals, missing test coverage, and stale assets. The business user sees the tables they touch most and the dashboards their team relies on. The landing page is persona-adaptive, automatically generated by an LLM, with widgets editable through plain-language instructions.

The chat now drives the navigation experience, not scrolling through a menu. Ask about failed services and Collate AI opens a side panel that summarizes what happened, why, and what to do. Learn about all pending approvals and receive a list of requests to batch review. Get data quality test recommendations pre-populated for your approval. The experience is designed to be tailored, actionable, and governed for every data persona to jump in quickly and get work done.

## Automations that pick their own agents

![Automations list with LLM-picked agents per task](https://cdn.hashnode.com/res/hashnode/image/upload/v1780604893681/adab03a9-1837-4c08-b28d-bb75846e2872.png)

New LLM-powered automations turn recurring work that fills a data team's week into a scheduled prompt. Describe an automation in plain language, like "scan the sales schema for null spikes every weekday at 9 AM and send the results to #data-quality," and Collate picks the right agent based on the prompt and the data sources. No agent selection dropdown. No manual wiring. The platform also ships templates for common patterns, such as Regulatory Compliance Tagging, Data Quality Playbook, and Lineage Impact Notifications. Eliminate manual toil for your data team with more intelligent automations.

## Collaborative AI chat

![Collaborative chat with humans and AI reasoning together over governed context](https://cdn.hashnode.com/res/hashnode/image/upload/v1780606678998/1f16a4e9-03f3-4bac-9557-ecdeee3841ee.png)

Bring together human teams and AI in one group chat, with multi-party reasoning over governed context. A steward opens a thread, the data engineer adds lineage, Collate AI pulls in the metric definition and the failing test. Everyone reasons in the same session, against the same governed metadata. Conversation threads support live presence, @mention notifications, role-based access, reactions, and a full audit trail of every contribution. When the thread lands on a correction, it can be saved as a memory nugget tagged to the asset, attributed to the user, and inherited by every future agent. Tribal knowledge stops disappearing into Slack screenshots, and becomes governed infrastructure that compounds with every conversation.

## Memory that compounds

![Memory list showing reusable nuggets tagged to data assets](https://cdn.hashnode.com/res/hashnode/image/upload/v1780606679852/c4ef1d86-fe55-493b-b52a-e8ebf01b1993.png)

Automatically capture preferences and tribal knowledge with a memory system shared across your data team and the AI agents working alongside them. Every correction a steward makes in a thread, every definition the team agrees on, every preference an analyst sets can be saved as a memory tagged to a data asset. These memories can be edited as needed, attributed to authors, tied to source conversations, secured by access controls, and tracked for usage. The next analyst can inherit those memories, as can the next AI agent. Memory stops being a lost Slack thread or a buried run log, and becomes a compounding moat as your data management gets smarter.

## AI Studio: Plugins and Skills

![Skill creation in AI Studio composing plugins and tools](https://cdn.hashnode.com/res/hashnode/image/upload/v1780606676770/e18e9bbc-8e57-4e79-8a09-50796d04f8cd.png)

Collate ships with agents out-of-the-box to handle the common data team work such as documentation, tiering, and classification. AI Studio is where teams can build their own agents, and 2.0 brings new capabilities for agent customization: **Plugins and Skills**.

Plugins connect agents to systems outside Collate over Model Context Protocol (MCP). Connect to key sources of context such as GitHub, Slack, Confluence, and Notion. Custom MCP servers connect through a MCP Plugin flow that auto-discovers the tools the server exposes: paste a URL, run Test Connection, get a populated tool list.

Skills compose plugins and tools into reusable, named capabilities. A skill bundles a plugin's tools with a trigger condition ("when should this skill be used") and an instruction ("what should the AI do using this plugin"). Once defined, a skill is callable from automations and chats. Each skill tracks where it is being used, so the steward who built "GDPR Policy Governance Skill" can see it running across two automations and three chats. For example, you can now easily connect to a Confluence page, analyze a GDPR policy located there without creating duplicate copies, and create instructions for how policy should be applied to your data.

## Open foundation, full audit trail

The whole experience runs on the open context layer of the open source OpenMetadata project, the same open foundation OpenAI chose for their internal data agent. Open industry standards sit beneath every action: JSON Schema for the metadata model, DCAT and DPROD for data product description, ODCS for data contracts, OpenLineage for cross-tool provenance, and Model Context Protocol for agent-to-system access. No proprietary metadata schema. No platform-specific lock-in. Your team's context inherits the openness, and so do the agents you build on top of it. Your data is strategic to your organization, and your context layer is just as important. Collate is built on open source and open standards to ensure your control and access to them.

Visibility into your context layer is just as crucial. Every action, by every actor (human or agent), lands in an AI Audit Log. AI-generated descriptions and corrections carry an AI-Generated Badge at the point of consumption, so a downstream user can see at a glance what an agent wrote and what a human verified. The audit trail is structured, queryable, and ready to satisfy the regulations that AI accountability now codifies: EU AI Act, DORA, BCBS 239, DPDPA, and GDPR. Compliance stops being a quarterly project, and becomes an ongoing function of how the platform works.

## Get started

The platforms that win in the AI era will be the ones built for both humans and agents to act inside together. Open foundations. Governed memory. Agents that inherit your team's context. This is what the data team's AI workflow finally looks like. Collate's new AI Native experience will ship as part of our 2.0 release.

[Request a demo](https://www.getcollate.io/contact-sales) to see Collate 2.0 running on your own data.

[#collate-2-0](/blog/tag/collate-2-0)[#ai-native](/blog/tag/ai-native)[#ai-for-data-teams](/blog/tag/ai-for-data-teams)[#openmetadata](/blog/tag/openmetadata)[#model-context-protocol](/blog/tag/model-context-protocol)[#ai-agents](/blog/tag/ai-agents)[#organizational-memory](/blog/tag/organizational-memory)

Ready for trusted intelligence?

See how Collate helps teams work smarter with trusted data

[Get Started](/welcome)[Contact Us](/contact-sales)

## Keep Reading

[View All](/blog)