Topic
AI / MCP readiness
Model Context Protocol (MCP) is the 2026 standard for how AI agents query data and trigger actions. For data pipelines, MCP-readiness means the vendor exposes an MCP server an agent can call to inspect pipeline state, run syncs, or pull recent data. The orchestrators are ahead of the ELT vendors here: Dagster+ lists an MCP server in preview on its plan comparison, and Prefect has built a managed MCP platform of its own. Nobody has turned MCP into a separate billing unit, and a vendor that briefly looked like it would has not.source - pulled 2026-06-21
Why MCP matters for pipelines
A typical AI agent that operates on warehouse data needs three things: discovery (what data exists), freshness (when was it last updated), and provenance (where did it come from). All three live in the pipeline tool. MCP gives the agent a structured way to ask. Pipelines without MCP support require custom REST wrappers or RAG-style indexes.
Vendor MCP support, September 2026
- › Dagster+: MCP server listed on the plan comparison as Preview, on every tier. The same table lists Dagster skills for Claude Code, Cursor and Codex as open source.source - pulled 2026-09-22
- › Prefect: ships FastMCP and sells Prefect Horizon, a managed MCP platform, as a product line alongside Prefect Cloud.source - pulled 2026-09-22
- › Airbyte: sells Airbyte Agents and a Context Store, but bills them through the ordinary credit meter. There is no separate agent billing unit on the rate card.source - pulled 2026-09-22
- › Fivetran: Roadmap item, no public MCP endpoint at June 2026.
- › dbt Cloud: Semantic Layer exposes a data-discovery API; MCP wrapper planned.
- › Estuary: CDC streams are agent-consumable via standard tooling.
Related
Written by Oliver Wakefield-Smith, Founder of Digital Signet. Independent reference, no vendor sponsorship.
Sources logged at /sources. Pricing-change history at /changelog. Last reviewed 2026-06-21.