San Francisco, October 27, 2026. MediaMelon today introduced Prisma, an AI agent that tells streaming teams why a problem happened, not just what went wrong and where. Prisma works from MediaMelon's complete, real-time data on every viewing session. With Hydrolix, it joins the affected sessions to the delivery logs that explain them. MediaMelon also opened its platform to customers' own AI agents through an MCP server, and works with AgileTV CDN Director to send each session to the CDN that performs best for it.
From what and where, to why
Most streaming problems show up first as numbers: more rebuffering, more failed starts, a drop in a quality score. The cause usually sits somewhere else, in a client build, a CDN, an origin or a home network. Prisma closes that gap. When a new app release starts failing on one device family, for example, Prisma scopes the affected sessions, reads the matching CDN request logs and returns a verdict: the build requests a path the origin never published. Every finding links to the raw requests behind it, so the fix goes to the right team with the proof attached.
"When something breaks during a live match, the team on call needs to know why, and who should fix it, while the match is still on. Prisma answers that with evidence. It joins the affected sessions to the delivery logs and links every finding to the raw requests. Evidence is what lets AI move from explaining a problem to helping resolve it, and that is where we are taking the product next: toward systems that recommend the fix and act on it, with a person in control. Everything Prisma uses is open through our MCP server, so our customers' own agents can build on the same data."
Built on every session, in real time
Prisma is only as good as the data under it. MediaMelon captures every viewing session on web, mobile and TV, not a sample, and makes it available in under 15 seconds. The platform is proven on AWS at more than 40 million concurrent viewers. MediaMelon operates it in its cloud or inside a customer's own cloud account, under the same service levels either way.
Open to your own agents
Everything MediaMelon computes is available through real-time data feeds, REST APIs and an MCP server, so a customer's own AI agents can query the same data Prisma uses. Ad and content teams can use Prisma, or their own agent, to build their weekly delivery and audience reports on demand. The AI layer is opt-in, the customer chooses where the model runs, and the model sees query results, not bulk viewer data.
Partners: logs and delivery decisions
Hydrolix. MediaMelon finds the sessions that went wrong. Hydrolix holds their logs. Prisma reads only the log lines it needs, with nothing bulk-ingested, and names the cause.
"Finally, you can bring your high-volume CDN logs together with your player-side analytics. See what logs lead to what outcome, without sampling or compromise."
AgileTV CDN Director. MediaMelon's viewer data informs which CDN serves each session. Every steering decision is compared with a random control group on the same traffic, so operators see what steering is worth on their own audience.
"CDN Director makes delivery decisions for every session, in real time. MediaMelon brings real viewer experience into those decisions, so business policies and quality of experience improve together. The control group shows our customers exactly what that is worth on their own traffic."
See Prisma at work in the two-minute demo, or book a demo on your own traffic.
About MediaMelon
MediaMelon provides real-time observability for the world's largest streaming services. Its platform measures every viewing session, content and ads, across web, mobile and TV, and turns it into answers for operations, ad and content teams. MediaMelon operates it for each customer, in MediaMelon's cloud or the customer's own. Based in San Francisco, MediaMelon works with media companies, broadcasters and operators across North America, Latin America, Europe and Asia.