Products & Tools

Clay's ChatGPT Plugin Now Builds Sales Workflows, Not Just Runs Them

Clay's relaunched MCP plugin builds and previews sales workflows inside ChatGPT, shipped alongside OpenAI's ChatGPT for Sales launch with 30-plus partner plugins.

Clay Adds Sales Workflow Building and Previews to ChatGPT Plugin
Clay Adds Sales Workflow Building and Previews to ChatGPT PluginPexels / pixabay
By James Calloway5 min read

Why it matters

  • Clay announced on November 17, 2026 that its relaunched MCP plugin builds and previews sales workflows directly inside ChatGPT.
  • The update shipped alongside OpenAI's ChatGPT for Sales, which includes a dedicated Sales plugin and more than 30 new or expanded partner plugins.
  • Workflows can draw on more than 200 data providers for prospecting and enrichment.
  • Building through the integration requires Editor or Admin access in the Clay workspace.
  • The MCP works on every Clay plan, including free and trial plans, using the same credits and actions as the Clay app.

Clay announced on November 17, 2026 that its relaunched MCP (Model Context Protocol, a standard that lets AI assistants call external tools) plugin can now construct and preview complete sales workflows directly inside ChatGPT. The update covers the full go-to-market chain: prospect selection, data enrichment and campaign launch. It marks a shift from Clay's earlier integration, which primarily let sales representatives run functions that teammates had already built and shared.

The timing matters. The announcement accompanied OpenAI's launch of ChatGPT for Sales, a product that combines a dedicated Sales plugin with more than 30 new or expanded partner plugins. OpenAI places Clay among its recommended tools for account prospecting and buyer research, including combining account signals with ChatGPT's web research to identify reasons to contact a prospect. In plain words: Clay is now part of OpenAI's official push into the sales software market.

What can users actually build in the chat?

The construction flow works from outcomes, not features. Users describe what they want, and Clay assembles a custom prospect audience, an enrichment workflow, a Sequencer campaign or a custom ad audience within the conversation. The plugin also selects which Clay features the requested outcome requires, so users can ask for a complete go-to-market play without first mapping that objective to individual features and settings themselves.

A typical workflow, according to the company, starts with a description of the ideal customer profile, the requested enrichments and the signals that could indicate a high-intent prospect. Clay then:

  • Returns prospects from the user's CRM or from its own audiences data
  • Builds workflows to find email addresses and phone numbers
  • Researches custom information with Claygent, Clay's AI research agent
  • Draws on more than 200 data providers for enrichment

Clay offered one concrete example. A user can create an audience of vice presidents of sales who are new in their roles, then add a workflow that automatically finds those prospects' companies' pricing pages. The example illustrates the integration's core claim: audience creation and subsequent research happen in a single chat-driven exercise, rather than as separate requests split across prospecting and enrichment stages.

The caveat is scale. Two hundred data providers is a supply-side number; the announcement does not say how accurate the generated workflows are, how often they need human correction, or how the system behaves with ambiguous prompts. The capability is new, and Clay has published no benchmarks for it. Takeaway: the building blocks are now conversational, but the quality of what gets built remains unmeasured.

How do previews and review work?

Interactive previews are the review mechanism for the generated work. ChatGPT can show a search before it runs it, refine the filters together with the user and display results as a table. It also presents workflows step by step, and shows campaigns before they go out. Clay's in-chat artifacts let users inspect what was built without opening the Clay application at all.

That detail addresses an obvious objection to AI-built sales machinery: if an assistant assembles an audience and launches a campaign, the user needs a way to check the work before anything reaches a real prospect. Previews are Clay's answer. They expose filters, workflow steps and campaign content inside the chat window, at the moment of construction rather than after the fact.

What the announcement does not say is equally notable. There is no mention of approval gates that block a campaign from sending without explicit human sign-off, only of previews the user can choose to inspect. Whether review is enforced or merely available is left open. In plain words: the inspection tools exist, but the responsibility to use them sits with the user.

Who can install it, and what does it cost?

Installation runs through ChatGPT's app directory. The sequence is: connect Clay, sign in, choose a workspace and authorize access, then enable Clay in a new chat. Existing users face a migration step. They must disable the earlier Clay for Reps MCP and install the new Clay MCP from OpenAI's marketplace.

Permissions are layered. Building through the integration requires Editor or Admin access in the Clay workspace, which means read-only roles cannot construct workflows from the chat. On ChatGPT Business, an administrator must enable Clay for the whole organization first, and each person's Clay workspace role then determines which tools that person can access. Partner-plugin availability also varies by ChatGPT plan, region and workspace settings.

The pricing position is straightforward. The MCP works on every Clay plan, including free and trial plans, and it uses the same credits and actions as equivalent work performed in the Clay app itself. There is no separate AI surcharge for building through ChatGPT. Clay added that users building in other tools can continue using its CLI (command-line interface), with updates to the MCPs for those tools expected in the coming weeks. No specific tools were named and no dates were given.

Why this matters

The expansion makes self-service construction the center of Clay's ChatGPT integration, and the company has threaded governance through the path from prompt to campaign: permissions determine who can build, previews expose what Clay produced, and the existing credit model accounts for the work done. That is a coherent design, and it aligns with how enterprises typically evaluate AI tooling — access control, visibility and cost accounting before adoption.

The stakes are real for the sales software market. OpenAI's ChatGPT for Sales, with its dedicated plugin and more than 30 partners, signals that conversational interfaces are becoming a distribution channel for go-to-market tools, not just a novelty layer on top of them. Vendors that plug in early gain placement; vendors that do not risk being intermediated.

For Clay, the unanswered questions are the operational ones. The announcement specifies no limits on audience size, no error rates for enrichment and no data-handling details beyond workspace-level authorization. Buyers evaluating the plugin for production prospecting will have to test those boundaries themselves. The coming weeks, when Clay ships MCP updates for its other building tools, will show whether this launch is a platform strategy or a single-channel bet.

Source: unite.ai

Share this article:

More from James Calloway

James Calloway

Show full bio

News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.

722 articles

Related articles

  1. OpenAI launches ChatGPT for Sales with 30-plus partner plugins
  2. OpenAI's 'Businesses in ChatGPT' Beta Puts AI Agents Between Shops and 1.2 Billion Weekly Users
  3. OpenAI Lets ChatGPT Users Buy Directly in Chat with Instant Checkout
  4. OpenAI Turns ChatGPT Into an App Store Challenger
  5. OpenAI opens ChatGPT app submissions and launches in-app directory

« Previous articleNext article »