Estuary
Agent Skills

Pipelines built by an agent, orchestrated by you.

Agent Skills lets your AI assistant create production-grade Estuary pipelines in the tool your team already uses. You describe the pipeline. The agent builds it. You review, approve, and run it.

  • No credit card required
  • 30-day free trial
You describe a pipeline, an agent builds the source-to-destination pipeline, and you review and approve itdescribebuildsreview & approveYousrcdestPipelineSKILL.mdAgent

Get to know Agent Skills

An open standard for agent-authored pipelines, production-grade from day one, with you in control of every change.

  • Built where you already work

    Your team lives in the assistant. The pipeline should get built there too.

    • Installs into Claude Code, Cursor, and the IDE of your choice
    • No new console, no tab-switching to remember a column name
    • SKILL.md is an open, portable standard that lives in your repo
    • Versionable like any other piece of infrastructure
    • No black box, no proprietary format, no lock-in
  • Production-grade by default

    Agent-built does not mean prototype-grade.

    • Runs on the same Estuary runtime that powers production today
    • Exactly-once delivery and schema evolution
    • Log-based CDC, streaming, and batch in one platform
    • Security posture for regulated workloads: SOC 2 Type II, HIPAA, GDPR, CCPA, CPRA
    • No separate "agent mode" with a different reliability story
  • Self-operating, not self-driving

    The agent handles the toil. You decide what runs.

    • Agent monitors the pipelines it builds and surfaces issues
    • Proposes connector-level fixes for schema drift and edge cases
    • You approve, version, and roll back
    • Every change is auditable
    • The agent does the work. The decisions stay with you.

Getting started

From a sentence in your assistant to a running pipeline, in three steps.

  1. 1

    Install the skill

    Add Agent Skills to your AI assistant. Works with Claude Code, Cursor, Codex, and 45+ other agents that support the open SKILL.md standard.

  2. 2

    Operate Estuary in plain language

    Ask the agent to build pipelines, modify captures and materializations, troubleshoot connector issues, inspect schemas, validate configs, prepare migrations, or make day-to-day operational changes. The agent uses the Estuary skill to understand the platform, gather the context it needs, and produce working configuration.

  3. 3

    Ship through your normal workflow

    The output is code in your repo, not a black box. Review the diff, approve the change, version it, and deploy. Estuary handles execution while your team keeps governance, review, and control.

You describe a pipeline in plain language, an Estuary Skill teaches the agent how to draft it, you review and approve, and Estuary deploys a Postgres capture to orders collection to Snowflake materialization pipeline.“Sync Postgres ordersto Snowflake”describedraftsiterateYouAgentSKILL.mdteaches the agenthow to use Estuaryreview & approvedeployLive data flowSourcecaptureorderscollectionDestinationmaterialization

Want to learn more about Skills? View docs →

See it in action

Watch a pipeline get built, reviewed, and shipped from inside the assistant.

Ready to build your first pipeline?

Free tier. No credit card. Install the skill and ship a pipeline from your assistant in minutes.

Try it Free

Frequently asked questions

    Which AI assistants work with Agent Skills?

    Agent Skills work with Claude Code, Cursor, OpenAI Codex, GitHub Copilot, Gemini CLI, and any other tool that supports the open SKILL.md standard. Claude Code users can install skills through the built-in plugin marketplace, and other tools can use the cross-tool Skills CLI or copy skills directly from the open-source estuary/agent-skills repository on GitHub.

    All of them. Dedicated skills cover the most common sources, including PostgreSQL, MySQL, MongoDB, and SQL Server CDC, plus HTTP webhook captures for tools like Shopify, Stripe, and GitHub. On the destination side, there are dedicated skills for Snowflake, BigQuery, Redshift, Databricks, and PostgreSQL. Beyond those, generic capture and materialization skills can configure any of Estuary’s 200+ connectors through dynamic schema discovery.

    They complement each other. Estuary’s MCP integration gives your assistant context about Estuary, so it can answer questions about the platform. Agent Skills take action: they provide a set of steps for the agent to run real Estuary workflows, from drafting and publishing pipeline configuration to checking task health, inspecting logs, and restarting connectors. Use MCP to learn about Estuary and Agent Skills to build and operate with it.

    No. The agent drafts pipeline configuration as code that you review before anything is published. You approve the change, version it like any other piece of infrastructure, and deploy it through your normal workflow. Every change the agent suggests stays reviewable, auditable, and reversible.

    Yes. Pipelines built through Agent Skills run on the same Estuary runtime as every other pipeline, with the same exactly-once delivery and schema evolution, using the same log-based CDC and optimized connectors. There is no separate agent mode with different reliability guarantees, and Estuary’s security and compliance posture applies to every pipeline regardless of how it was created.

    Create a free Estuary account, install the flowctl CLI, and add the skills to your assistant. Claude Code users can run /plugin marketplace add estuary/agent-skills, and other tools can use npx skills add estuary/agent-skills. From there, ask your assistant something like “capture my Postgres database into Estuary” and it will guide you through setup.