Skip to main content
Community Hub

Atlan products

TL;DR

Atlan offers two products—the Agentic Data Catalog for governed context about your data estate, and Atlan Trident for registering, observing, and governing your AI agents and skills. Learn what each one does and which one you need.

Your AI can read this via Docs MCPcurl -fsSL "https://docs.atlan.com/install-docs-mcp" | bashConnect

AI needs two kinds of context to be useful in your organization: context about your data, and context about your agents. Atlan builds both.

Agentic Data Catalog gives AI agents the governed understanding of your data they need to answer business questions correctly—what tables mean, how data flows, who can use what, and what your organization has decided about quality and trust.

Atlan Trident gives your organization a governed home for the agents themselves—where they live, what skills they run, what they did, what they cost, and whether they're getting more accurate over time.

Most organizations need both. The two products connect: an agent built on Trident reads governed context from the Agentic Data Catalog through MCP, and its traces flow back into Trident so you can measure and improve it.


Agentic Data Catalog​

The Agentic Data Catalog is Atlan's data intelligence platform. It connects your data estate, turns raw metadata into governed business knowledge, and serves that knowledge to your team and your AI agents.

What it does​

When you connect a data source to Atlan, metadata flows in automatically: table names, column types, ownership, usage, freshness, and lineage. Atlan ties this together into your Enterprise Data Graph—a living map of everything in your data estate and how it all connects.

On top of that graph, your team adds meaning: business terms, ownership, certifications, quality rules, data contracts, and policies. That human context is what makes data trustworthy and AI-ready.

How AI uses it​

AI agents are only as good as the context they have. An agent that doesn't know what customer_mrr means, who owns it, or whether it's been certified gets the numbers wrong.

The Agentic Data Catalog solves this at three levels:

  • Atlan MCP—expose governed metadata from your entire data estate to any AI tool that supports MCP: Claude, Cursor, Copilot, your own agents. Agents read your actual business definitions and lineage instead of guessing.
  • Context Agents Studio—run AI agents inside Atlan that enrich your catalog at scale: writing descriptions, tagging columns, filling gaps in documentation across thousands of assets.
  • Context Engineering Studio—build a semantic model from your governed context and deploy it to Snowflake, Databricks, or dbt so analytics agents answer questions with the metrics your team actually agrees on.

Who it's for​

  • Data and analytics teams that need a catalog to discover, understand, and trust data
  • Governance and compliance teams that need to tag sensitive data, enforce access policies, and prove lineage
  • AI and platform teams that need to give agents governed, business-ready context about the data estate

To get started, see What is Atlan? or the quick start for admins.


Atlan Trident​

Atlan Trident is the AI registry for your organization. It's where your agents and the skills they use live, where their work is recorded, and where you control what they can reach.

What it does​

As AI agents multiply across teams and tools, two things break down: visibility and governance. You can't answer basic questions—which agents exist, what do they actually do, what do they cost, are they getting better or worse? And you can't control what models and tools they're allowed to reach.

Trident solves both. It gives every agent a home with a full record of its skills, versions, sessions, traces, and evals. And it routes model and tool calls through AI Gateway so you control access without touching the agents themselves.

Key capabilities​

  • Skill registry—publish versioned skills from GitHub or the CLI. Every skill has an owner, a version history, and a record of which agents run it.
  • Agent registry—register agents with their skills, models, and sessions. See every run, its inputs, outputs, cost, and whether it succeeded.
  • Traces and evals—collect execution traces automatically. Build a golden set of questions with known-good answers and run evals to measure accuracy over time—so you know whether a change made the agent better or worse.
  • Metric Studio: define your organization's metrics and trusted SQL in one place, then give analytics agents access to them. Talk-to-data agents answer business questions with the numbers your team agrees on, not hallucinated approximations.
  • AI Gateway—provide multi-provider LLM routing, rate limiting, per-tenant budget controls, cost tracking, and observability. Manage which MCP servers your agents can reach. Keep data within private network boundaries.

Who it's for​

  • Platform and AI engineering teams that build, share, and run agents across the organization
  • Analytics and data science teams that want talk-to-data agents to answer questions accurately, with measurable improvement over time
  • Security and compliance teams that need visibility and control over what AI agents can access and do
note

Atlan Trident is in private preview. See Availability and limits for what's on by default. Atlan Trident was previously called AI Registry and Agent Registry.

Documentation lives at platform.atlan.com.


How two products work together​

The Agentic Data Catalog and Atlan Trident are designed to be used together, though each works independently.

They're separate products. Each has its own sign-in and its own MCP server: the Agentic Data Catalog's is at https://mcp.atlan.com/mcp, and Atlan Trident's is at https://api.atlan.com/mcp. Atlan Trident doesn't import catalog assets, lineage, or glossary terms; agents reach the catalog by adding its MCP server to Atlan Trident's MCP Gateway. For what each product does and how they connect, see Why use Agentic Data Catalog with Atlan Trident.

A typical AI workflow looks like this:

  1. Context from the catalog—an agent reads your governed data context through Atlan MCP: table definitions, column meanings, certified metrics, lineage, and ownership.
  2. Skills in Trident—the agent runs a published skill from Trident's skill registry, with a known version and owner.
  3. Gateway enforcement—the model call passes through AI Gateway, which applies your model and tool access policies.
  4. Traces back to Trident—the execution trace is recorded in Trident. You can see exactly what the agent did, what it read, and what it returned.
  5. Evals improve accuracy—the trace becomes an input to your eval suite. If the agent answered a metric question wrong, you update the semantic model in the catalog and re-run.

This loop—governed context in, observable traces out, evals closing the loop—is how you take an analytics agent from a demo to production.


Open source from Atlan​

Atlan also publishes two open-source projects:

  • Registry OSS—an open specification for agent registries and aireg, a CLI that works with any registry that follows it (Apache 2.0). Atlan Trident is the hosted product built on this standard, adding workspace permissions, SSO, approvals, traces, and evals.
  • Analytics agents playbook—a practical method for building talk-to-data agents that get business numbers right: metrics from analyst SQL, a golden question set, and improvement loops. Free to reuse under CC BY 4.0.

See also​