MCP Advertising Agents: Inside Omneky's New Public API

Photo by Szabo Viktor on Unsplash

One URL. Zero design briefs. Four finished ad formats.

That is the contract Omneky made public on July 10, 2026, when the San Francisco-based ad creative platform opened its generation engine to developers, commerce platforms, and AI agents for the first time. According to a press release distributed by PR Newswire, the company launched both a public REST API and a Model Context Protocol (MCP) server — the same open standard that Amazon Ads, TikTok, Google, and Meta have each shipped in the past year. The infrastructure shift is quietly significant: the ad creative production bottleneck just became an API call, and every major ad platform now speaks the same agent protocol to reach it.

What Shipped

Omneky's API accepts a brand URL and returns completed creatives in 1:1, 4:5, 9:16, and 16:9 aspect ratios — the four formats that cover Meta feeds, Reels, Stories, and YouTube pre-roll in a single pass. The system reads brand positioning, extracts color palettes and fonts, writes headlines, and generates calls-to-action without human input or intermediate prompt engineering. The MCP server layer extends this further: AI agents like Claude can trigger full multi-format creative generation through conversational commands, eliminating the dashboard entirely.

Omneky was founded in 2018 by Hikari Senju, a Harvard computer science graduate who previously built and sold a tutoring app to Yup.com. The company has raised $11.45 million from investors including SoftBank, AIX Ventures, and Village Global, with a seed round accounting for $10 million of that total. Senju framed the API launch in access terms: "Omneky's mission is to democratize growth — to make sure the next great founder doesn't need a million-dollar ad budget to reach the people who'd love what they've made."

The timing is deliberate. TikTok shipped its Ads MCP server in May 2026, becoming the last of the four major ad platforms to standardize agent access. Google upgraded its Asset Studio with multimodal Gemini-powered creative generation in 2026. Amazon opened its Amazon Ads MCP Server to open beta the same year. MCP — the open standard Anthropic introduced in November 2024 — has now become the connective tissue the advertising stack was missing.

The Pattern: MCP as Advertising Infrastructure

Mercury Media Technology's analysis of MCP adoption named the structural problem clearly: "MCP solves a problem that's been limiting AI adoption in digital advertising: every connection between agentic AI systems and external platforms requires custom integrations. MCP standardises those connections." That is the pattern here — not a new AI creative tool, but a standardization event.

Before MCP, building an autonomous campaign workflow meant maintaining brittle point-to-point integrations: one connector for Meta's API, another for Google Ads, a third for TikTok, a fourth for the creative layer. Each required separate authentication flows, separate error handling, and separate maintenance cycles. With all four major ad platforms now shipping MCP servers alongside third-party creative providers like Omneky, an AI agent can orchestrate a full campaign lifecycle — creative generation, placement selection, launch, and optimization — through a single protocol layer.

This is the same tool-use orchestration pattern that shows up across agentic AI: the agent does not generate output itself, it sequences API calls to specialized tools and assembles results. What has changed in advertising is that the tool endpoints now speak a common language. As noted in a related analysis of how Claude Code burned through Uber's entire AI budget, the cost architecture of agentic systems matters enormously — and MCP's standardization directly reduces the number of bespoke tool calls an agent needs to make to accomplish a task, which is the primary lever on token expenditure at scale.

As of July 11, 2026, according to GlobeNewswire market research, the AI in advertising market stood at $11.17 billion in 2025 and is on track to reach $14.12 billion in 2026, growing at a CAGR of 26.4%, with a projected $36.34 billion by 2030. The Omneky launch lands squarely in the middle of that acceleration curve.

Chart: AI in advertising market size — $11.17B (2025) to a projected $36.34B by 2030, per GlobeNewswire market research as of July 2026. The 26.4% CAGR makes this one of the fastest-growing enterprise AI categories.

Photo by Nangialai Stoman on Unsplash

What the Implementation Actually Looks Like

Strip away the press release language and the Omneky API follows a recognizable agentic tool-use pattern with three discrete stages:

1. Input normalization: The agent passes a brand URL. Omneky's system fetches and parses the page, extracting visual identity tokens (colors, typography) and semantic positioning signals (category, value proposition, tone of voice).
2. Creative synthesis: A multimodal generation layer combines those extracted brand signals with performance-optimized layout templates to produce platform-ready creatives across all four aspect ratios simultaneously — no intermediate human approval step.
3. MCP tool response: Through the MCP server, the entire output surfaces in the agent's context as structured tool-call results — file references, metadata, format dimensions — ready for the next workflow step: ad platform upload, A/B variant branching, or a human review gate if the team wants one.

For comparison context, Smartly's AI Studio generated 1.9 million creative assets across 260+ customers, with self-reported 30x faster production and a 27% performance lift versus static creative. The key architectural difference: Smartly operates as a managed platform with its own orchestration layer, while Omneky's public API and MCP server expose the creative layer as a composable service any agent can call independently. For teams already building agentic ad workflows, that composability matters more than the managed-platform alternative.

Pricing context is also relevant here. As of July 2026, most AI ad generators run between $19 and $99 per month for small ecommerce brands, compared to $500–$3,000 per campaign at traditional agencies. Omneky has not published public API pricing tiers, but the cost envelope for API-driven creative generation sits well below the agency alternative for brands running continuous campaign cycles.

Where This Breaks in Production

The demo always looks clean. The production deployment rarely is.

The most concrete failure mode in the available data: AI-generated creatives achieve 12% higher CTR on Meta than manually produced ones across 50,000+ ad variations, but show a -8% conversion gap for high-AOV (high average order value) products. That delta matters enormously for DTC brands selling premium items. Higher click-through with lower purchase completion means paying for traffic that does not convert — a real risk when the creative layer is fully autonomous and lacks the nuanced judgment a human art director applies to luxury or considered-purchase categories. Eval-driven development that tracks downstream conversion by product segment, not just CTR, is the mitigation here.

Second failure mode: brand extraction from a URL is a best-guess inference, not a controlled input. If a brand's website has inconsistent visual identity, outdated copy, or a slow-loading homepage, the extraction step inherits those problems and propagates them into every generated creative across four aspect ratios simultaneously. Garbage in, garbage out — at four-format scale.

Third: MCP tool-call loops. When an AI agent is tasked with generating and launching a campaign, the creative generation step might succeed while the ad platform upload step fails, leaving the agent without clear state on what was actually deployed. Without deterministic state management between tool calls, autonomous campaign agents are prone to partial-execution states that are difficult to detect and harder to roll back. This is a general problem with multi-step agentic workflows — not specific to Omneky — but worth flagging before any team assumes "autonomous" implies "self-healing."

Fourth: context window blowup risk. If an agent is managing multiple simultaneous campaigns and each creative generation call returns structured metadata for four ad formats, cumulative tool-call output can saturate available context before the agent reaches the optimization loop. Teams building on Omneky's MCP server should plan for result pagination and explicit context management from the architecture stage, not as an afterthought.

Who Should Move Now — and Who Should Wait

Move now if you are a developer building agentic ad infrastructure, a performance marketing team running high-volume campaigns on standardized brand guidelines, or an ecommerce platform that wants creative generation embedded in merchant onboarding flows. The URL-in, four-formats-out contract is genuinely useful and the MCP interface means you are building to a standard all four major ad platforms now support.

Wait if your brand operates in high-AOV categories where the conversion gap is a real cost, if your visual identity is inconsistent or lives primarily in offline assets, or if you do not yet have state management and error handling built into your agent workflows. Deploying autonomous ad creation without those guardrails does not save time — it makes failures faster and harder to attribute.

As of July 11, 2026, 61.4% of marketers have incorporated AI into their marketing strategies. The infrastructure to make that AI fully autonomous — covering creative, placement, and optimization — is now largely in place across the major platforms. In my read, the Omneky launch is less about any single company's capabilities and more about a threshold moment: when all four major ad platforms and at least one credible third-party creative layer share a common protocol, the agentic advertising workflow stops being a custom engineering project and starts being a configuration problem. That is a genuine shift — and the failure modes above are exactly what separates teams that capture the gains from teams that learn about them the hard way.

Frequently Asked Questions

Disclaimer: This article is for informational and educational purposes only and does not constitute marketing, financial, or business advice. All statistics are sourced from publicly available research and press materials. Research based on publicly available sources current as of July 11, 2026.