AppsFlyer announced a significant expansion of its agentic AI offering. The release gives marketing teams the architecture they need to run AI agents reliably and transparently, with marketers in control. It extends AppsFlyer Agent Hub so marketers can build their own agents and introduces the company’s first agent that can take action in an advertising platform. It also adds four new Model Context Protocol (MCP) capabilities, so teams can work with fraud, creative, campaign link and integration data directly from their own AI tools.
Through 2026, the challenge for many marketing teams adopting agentic AI has shifted from what agents can do to how much visibility and control teams have over their actions. AppsFlyer is addressing this through governed agentic AI infrastructure. With transparent decision-making accessible in every workflow, teams can understand, evaluate, and trust the reasoning behind every AI recommendation. Marketers will be empowered to strategise based on recommendations from agents, whether that is investigating fraud spikes, spotting creative fatigue, or recommending Apple Ads bid changes.
“Marketing teams are moving from manual campaign orchestration to agentic execution,” said Inna Weiner, VP Product at AppsFlyer. “With this release, AppsFlyer is giving marketers trusted, agentic tools that can analyse, recommend and act on their data to deliver clear value for growth.”
Agent Hub: From insight to approved action
With the expanded Agent Hub, marketers can create their own agent by describing any marketing performance workflow. The agent pulls the relevant data down to event level, analyses it and delivers insights to Slack or email on a schedule the team sets. It can also run root cause analysis and recommend the next action.
The release also introduces the Apple Ads Share of Voice Agent, the first AppsFlyer agent that goes beyond recommending to acting. It estimates share of voice for each keyword and geography, recommends a bid change when share of voice is off target, and applies the change in Apple Ads directly from AppsFlyer. Updates are applied only on the marketer’s approval, so they keep full control. The agent is in closed beta and available through AppsFlyer customer success managers.
MCP: AppsFlyer data inside marketers’ own AI tools
AppsFlyer’s new MCP tools bring more of the platform into marketers’ own AI tools. Protect360 fraud data is now available through MCP, so teams can, for example, break down fraud concentration by geography over the last 30 days, or find out which custom validation rules drove the most blocks in a given period. Creative performance data is now exposed at creative level, covering spend, reach, conversion, retention and revenue cohorts. That helps teams identify winning and fatiguing creatives faster and build automation around them.
Two further capabilities extend this to campaign setup and support. OneLink Link Creation turns a plain-language campaign description into a fully configured OneLink, and nothing is created until the marketer approves it. Partner Integration Settings surfaces a partner integration’s configuration, attribution windows, postback rules and event mappings to any MCP client, so marketers and support teams can diagnose setup issues in conversation. All MCP capabilities are available in the AI assistant.
“Few markets have embraced AI with the conviction of the Gulf,” said Sarah Maina, Regional Manager, Middle East & France, at AppsFlyer. “That ambition is especially visible across the region’s fast-growing gaming, fintech and e-commerce app sectors, where marketing teams are eager to put agents to work. What they need now is the confidence that they can understand the reasoning behind every recommendation and approve actions before they happen. This release gives regional marketers that foundation, so they can scale agentic AI responsibly and in step with the region’s broader AI ambitions.”
Discussion about this post