This list is for performance marketers, media buyers, and agency leads who already do their research in Perplexity and want the same chat to touch real ad accounts. The tools were ranked on four things: how well they connect to Perplexity through connectors or MCP, whether they act or only report, which ad platforms they cover, and how much setup they demand. MCP (Model Context Protocol) is an open standard that lets AI apps call external tools, and a connector is the way Perplexity attaches to one. Support depends on plan and platform as of November 2026, so check Perplexity's current help docs for setup steps, and confirm each vendor's compatibility, features, and pricing on its own site before you commit.
Quick Comparison of the Ad Agents and MCP Connectors
- AdStellar: Meta advertisers and agencies who want Perplexity to launch campaigns; from $100; builds and launches full data-backed Meta campaigns from chat.
- Pipeboard Meta Ads MCP: DIY Meta buyers who want a minimal connector; see website for pricing; a lightweight Meta-only bridge for any MCP client.
- Google Ads MCP Server: technical Google Ads teams who can self-host; free; first-party open-source code from the platform owner.
- Supermetrics MCP: teams comparing performance across many channels; see website for pricing; the broadest cross-channel data coverage here.
- Windsor.ai MCP: analysts asking attribution questions; see website for pricing; attribution data and ad-platform data in one connection.
- Adzviser: small agencies needing fast client-ready answers; see website for pricing; no-code client reporting plus chat from one account connection.
- Zapier MCP: ops-minded marketers automating cross-tool workflows; see website for pricing; chains research into actions across thousands of non-ad apps.
- Composio: technical teams building a custom ad agent; see website for pricing; managed authentication and tool routing for bespoke agents.
1. AdStellar
AdStellar is an AI media buyer for Meta Ads (Facebook and Instagram). It analyzes your past performance, builds campaigns, generates the creatives, and launches them. It connects to Perplexity through MCP, and it is also available as an approved ChatGPT app, so the same workflow works in either assistant.
The distinction on this list is execution. Most entries below hand Perplexity data to read. AdStellar gives it an agent that acts: ask for a campaign in chat and it assembles a full, data-backed structure from your account history, explains the reasoning behind it, and launches it. Setup is short. In Perplexity's connector settings you add a custom remote MCP connector using the URL https://app.adstellar.ai/api/mcp, then authorize your account. Because connector availability varies by plan and platform as of November 2026, confirm the exact steps in Perplexity's help docs. ChatGPT users can use the approved app at this link.
- A conversational Agent analyzes your account, creates campaigns, and launches them, so one chat replaces several tabs and handoffs.
- Seven specialized AI agents each handle part of the build, and the reasoning behind each campaign is shown, so you can audit decisions instead of trusting a black box.
- AI image ads, video ads, UGC avatars, and talking-head ads, plus Product Video, Ad Clone, and URL to Ad Maker, remove the wait for creative production.
- Bulk launch pushes hundreds of ad variations live in minutes, which makes real creative testing practical.
- Element-level creative scoring, fatigue detection, and a ranked winners library show which headlines, images, and hooks work and let you reuse them.
It also imports existing Meta campaigns, offers unlimited workspaces for agencies managing many accounts, and integrates with Cometly for attribution. A marketer or media buyer runs it day to day, and no engineering is required.
The limitation is scope. AdStellar covers Meta only, so it does not suit Google-first or TikTok-first advertisers, and you would pair it with another tool for cross-channel reporting. Pricing starts at $100 and runs up to $10,000 across plans; confirm current tiers on the website.
Best for: Meta advertisers and agencies who want Perplexity to launch campaigns, not just report.
2. Pipeboard Meta Ads MCP
Pipeboard offers a third-party MCP server that lets AI clients read and manage Meta Ads accounts in conversation. It is the closest like-for-like Meta alternative to AdStellar, but it is a bridge rather than a buyer.
What it does well is stay small. You connect a Meta ad account, then ask Perplexity or another MCP client questions about campaigns, ad sets, and ads in plain language. A hosted remote connection is offered, though you should verify current details and how it behaves in Perplexity before relying on it.
- Meta ad account access over MCP means your own assistant can query live account data.
- Campaign, ad set, and ad retrieval gives you the structure and numbers behind any question.
- Natural-language performance questions replace manual Ads Manager exports.
- A hosted remote connection (verify current details) can cut the work of running your own server.
The tradeoff is that the intelligence is yours. There is no creative generation and no built-in buying strategy, so the quality of recommendations and any changes depends on the prompts you write and the judgment you apply. That suits a buyer who already knows what they want to ask, and it frustrates one hoping for a system that proposes a plan. Meta only, like AdStellar. Pricing and plan limits are listed on the vendor's site, so check there.
Best for: DIY Meta buyers who want a minimal connector into their own prompts.
3. Google Ads MCP Server
The Google Ads MCP Server is an open-source project published in Google's Google Ads developer repository. It lets AI assistants query Google Ads account data through the Google Ads API.
Its edge is provenance and cost. It is first-party code from the platform owner, it is free, and you can read exactly what it does. For a Google-first team, that transparency counts for a lot.
Setup is the catch, and it is not trivial. You need a Google Ads developer token, API credentials, and somewhere to run the server, which makes this an engineering task rather than a five-minute connector. It is not a polished hosted product, and Perplexity compatibility should be verified, since MCP clients differ in how they handle local versus remote servers.
- Natural-language querying of Google Ads data turns account questions into API calls for you.
- Open-source code lets your developers inspect, audit, and extend it.
- Built on the Google Ads API, it reads the same data your reports use.
- Compatibility with MCP clients means it is not tied to a single assistant.
It is oriented toward querying, so do not expect campaign building, creative generation, or strategy. Cost is free, but API access requirements apply and your own hosting and developer time are the real price.
Best for: Technical Google Ads teams comfortable self-hosting.
4. Supermetrics MCP
Supermetrics extends its long-established marketing data connectors to AI assistants through MCP. If your question spans platforms, this is the data layer that already knows how to join them.
Coverage is the differentiator. Meta Ads, Google Ads, TikTok Ads, and Google Analytics sit behind one connection, with metrics unified across them. That makes questions like "which channel delivered the cheapest qualified traffic last month" answerable in one prompt, which no single-platform bridge on this list can do.
- Cross-channel ad and analytics data in one place saves stitching together separate exports.
- Unified metrics make numbers comparable across platforms with different definitions.
- Natural-language data questions let non-analysts get answers from Perplexity.
- An established connector library means common sources are already supported and maintained.
The limitation is that it is analysis-oriented. It does not create or launch campaigns, so it informs a decision that you or another tool must then execute. A common pattern is pairing it with an execution agent. Setup is moderate: authorize your data sources, then attach the connector in your AI client, after verifying how it works with Perplexity. Pricing is listed on the vendor's site and typically depends on which connectors and usage you need, so confirm current plans.
Best for: Teams comparing performance across many ad channels.
5. Windsor.ai MCP
Windsor.ai exposes its connected marketing and attribution data to AI clients through an MCP server. Where Supermetrics leads on breadth of channels, Windsor leans toward attribution.
The value is having attribution data sit next to ad-platform data in one connection. A marketer can ask Perplexity how Meta and Google spend compare against attributed conversions from GA4, rather than trusting each platform's self-reported numbers in isolation. That framing matters when platform dashboards disagree.
- Marketing data integration pulls many sources into one queryable layer.
- Attribution modeling data lets you question channel credit, not just platform clicks.
- Natural-language querying removes the need to build a report before asking.
- Many source connectors mean you can add channels without new tooling.
It is read and analysis only. There is no ad creation and no launching, so it will not close the loop from insight to action. It also assumes your tracking is in decent shape, because attribution answers are only as good as the data feeding them. An analyst typically owns setup and day-to-day use. Verify Perplexity compatibility and current plan details on the vendor's site before you buy.
Best for: Analysts asking attribution-flavored channel questions.
6. Adzviser
Adzviser is a reporting tool that connects ad accounts to AI chat for quick questions and no-code dashboards. It is built around the agency habit of needing an answer a client can read.
The standout is combining two jobs from one account connection: ask questions in chat, and produce dashboards and summaries without building them by hand. For a small agency juggling many clients, that shortens the loop between a question and something presentable.
- Multi-platform ad account connection covers Google Ads and Meta Ads from one place.
- AI chat Q&A on ad data answers quick questions without opening each platform.
- No-code reporting dashboards let account managers build views without an analyst.
- Client-ready summaries cut the time spent turning numbers into explanations.
It analyzes rather than executes, so it will not build or launch campaigns, and it also integrates with Google Sheets for teams that live in spreadsheets. Perplexity support should be verified, since some of its chat experience may run inside its own interface or other assistants. Setup is light and suits a non-technical account manager. Pricing is on the vendor's site, so check current plans.
Best for: Small agencies needing fast client-ready answers.
7. Zapier MCP
Zapier MCP exposes Zapier's app actions to MCP-compatible AI clients, which turns a chat into a trigger for automation across tools.
It is the odd one out here because it is not an ad tool at all. Its strength is chaining. A Perplexity research session can end with a row added to Google Sheets, a Slack alert to the team, or a HubSpot record created, all configured without code. For ops-minded marketers, that glue is often the missing piece.
- Thousands of app actions give you reach well beyond advertising.
- Workflow automation from chat lets one prompt kick off multi-step processes.
- Spreadsheet, CRM, and Slack actions cover the common reporting and handoff tasks.
- No-code configuration means a marketer can set it up without a developer.
The limitation is depth. Ad-platform actions are shallower than a purpose-built media buyer, so expect basic operations rather than strategy-driven campaign construction. Calls also count against your plan's usage, so a chatty workflow can burn through allowances. A marketing operations person usually owns it. Pricing depends on plan tier and task or call volume, so check the vendor's site for current numbers.
Best for: Ops-minded marketers automating cross-tool workflows.
8. Composio
Composio is a developer platform that provides managed authentication and tool access for building custom AI agents. It is infrastructure, not a finished ad agent.
Its distinctive job is solving the unglamorous parts of agent building. Managed OAuth handles the credentials for each connected app, and tool routing gives your agent a catalog to draw from, with MCP server support and SDKs for custom builds. If you want an ad agent shaped around your own process rather than someone else's, this removes a lot of plumbing.
- Managed OAuth for agent tools spares you from building and maintaining auth flows.
- A large toolkit catalog gives your agent many apps to act on.
- MCP server support lets your build plug into MCP-compatible clients.
- SDKs for custom agents let engineers control logic, prompts, and guardrails.
The honest limitation is effort. You need engineers, and ad-platform toolkit coverage must be verified for the exact actions you need, because a long catalog does not guarantee deep support for Meta or Google campaign management. You also own the strategy layer, evaluation, and ongoing maintenance. Pricing is listed on the vendor's site and generally scales with usage, so confirm current terms.
Best for: Technical teams building a custom ad agent.
Matching the Right Connector to Your Ad Workflow
The main dividing line on this list is between tools that act and tools that report. A read-only MCP server can answer questions about your account, but it is not an ad agent, and an MCP server will not necessarily behave the same way in every AI client. Test any connector inside Perplexity itself before building a process around it.
- Meta advertisers who want Perplexity to launch campaigns: AdStellar. It builds, creates, and launches from chat, and it reaches Perplexity at https://app.adstellar.ai/api/mcp.
- Cross-channel analysis: Supermetrics for the widest channel coverage, or Windsor.ai when attribution is the question.
- Technical Google Ads teams: the Google Ads MCP Server, if you are comfortable self-hosting.
- Custom workflows: Zapier for no-code automation, or Composio when engineers are building a bespoke agent.
Many teams will combine two of these, such as an analysis layer for cross-channel questions and AdStellar for execution on Meta. If Meta is where your budget goes, start there and see what a chat-driven media buyer does with your own account history. Learn more about our services.
