What is Relay.app doing strategically?

In July 2025, Relay shipped bidirectional MCP support: teams can expose Relay workflows as callable tools in Claude and Cursor, and Agent steps in Relay can call external MCP servers. The feature is documented publicly in product docs and the July 2025 changelog.

MCP is becoming the de facto integration layer for agentic AI tools. A Relay workflow that is callable from Claude is effectively embedded in any AI coding or ops stack that uses Claude as the reasoning layer. For RPA and Browser Automation founders, this means Relay can occupy the action execution slot in an AI stack without the buyer needing to visit relay.app to get value.

This is a real distribution move, not a marketing checkbox. The constraint is that remote MCP server connections currently require Anthropic Claude models, which limits breadth. If Relay extends MCP to other model providers, the moat deepens materially.

Relay shipped native Tables in November 2025, publicly announced in the January 2026 product update. Teams can now store, query, and manage structured data directly inside workflows without relying on Google Sheets or Airtable.

Every external tool dependency is a switching cost going the wrong way: it is a reason to leave, not stay. Pulling data storage into the platform removes the most common integration a mid-market ops team uses alongside Relay. When data lives inside Relay, the audit trail, the workflow logic, and the data are all in one renewal conversation.

This is the kind of feature that does not win competitive deals but prevents churn from existing customers. For competitors, it raises the bar: you now need to match not just workflow capability but embedded data persistence to dislodge Relay from an ops team that has been running for three to six months.

Relay meters usage on both workflow steps and AI credits. All integrations are unlocked at every tier, which simplifies the upsell conversation. But AI-heavy actions compound fast: profile scraping costs 7 credits, image generation costs 50, and a 100-run workflow combining scraping and summarization can exhaust the monthly credit allowance in days. Credit add-ons scale from $11 per month for 5,000 credits up to $600 per month for 200,000 credits.

The all-integrations-included framing is a strong acquisition message, but the credit model creates unpredictable cost at scale for any team running high-volume AI steps. Buyers who discover the ceiling mid-month face a forced upsell decision, which is a real retention risk and a clear positioning opportunity for competitors who can offer flatter or more transparent pricing for equivalent AI automation volume.

Relay has made a deliberate tradeoff: simplify the feature story, complicate the cost story at scale. For a Founder or CTO building in this space, this is the most actionable pricing wedge available right now.