Adobe goes all-in on agentic AI and opens Creative Cloud to outside agents via MCP

Adobe puts agents across the creative suite

At its MAX conference in Miami, Adobe announced a company-wide push into agentic AI: new AI features across Photoshop, Premiere, Illustrator and InDesign, an upgraded Firefly AI Assistant, always-on background agents, and a local Model Context Protocol (MCP) integration that lets outside AI agents edit Creative Cloud projects directly.

According to The Verge, the local MCP integrations — previewed rather than shipped — are designed to give users a choice between Adobe's own assistants and third-party agents from providers such as OpenAI, Google and Anthropic. Those external agents would run locally in Photoshop, Illustrator, Premiere and After Effects and make direct changes to native project files. Because all layers, vectors, bins and sequences remain editable, a human can take over at any point to refine the work. Agents connected to other MCP-capable tools could also span applications, for instance pulling reference material from Google Docs, building a design in Adobe's tools and posting the result to Slack.

The two reports differ on timing. The Verge says the Creative Cloud MCP integrations are not available yet and that a private beta is planned for early 2027, while TechCrunch describes Adobe as releasing a local MCP that exposes endpoints across Illustrator, Photoshop, After Effects and Premiere. TechCrunch quotes Adobe vice president Deepa Subramaniam saying the company is deliberately starting open-ended: "We don't have any need or desire to sort of limit what the MCP endpoint can expose."

Always-on agents and a smarter Firefly assistant

The Verge also reports a preview of "always-on" agents that run recurring jobs continuously in the background — organizing incoming assets, running quality checks or preparing content — while the user works on something else or steps away entirely. Users assign the job, review the agent's plan and approve the final output; one example is automatically turning incoming images into promotional graphics for multiple social channels. Adobe has not set a launch date, but says the capability will arrive in the Firefly AI Assistant first. TechCrunch frames this as Adobe making its agent proactive and always on, so it can gather context about the user's work and be present for any action.

The Firefly AI Assistant beta itself gains a Plan Mode for reviewing and editing an agent's plan before it executes a task, a Model Selector that lets users pick a specific AI model or leave it on auto-select, and an integration with Firefly Boards. Users can also set preferred formats, colors, brand guidelines, characters and products, and have those preferences applied automatically across every project.

Since its introduction earlier this year, Adobe's creative agent has appeared in Premiere, Photoshop, Illustrator, InDesign, After Effects and Frame.io, per The Verge. It is now pushing deeper into video: in Premiere it can track masked subjects through a frame and apply effects or adjust color and exposure across entire clip sequences, and it is coming to the Premiere mobile app for conversational edits on the go.

The rest of the MAX announcements

TechCrunch rounds out the picture on individual apps: Photoshop gets a relighting feature and sliders for prompt-based generative styles; Premiere gains a professional color grading system and an object removal tool for video; Lightroom adds blinking-eye correction and blemish removal. In Firefly, Adobe is launching Builder, a prototyping tool in private beta that pushes the app into Figma's territory, alongside a tool for on-brand graphs, an image editor that changes part of an image while leaving other objects untouched, and an AI video tool that assembles clips, images and sound effects into a finished video. Quick actions such as background removal and video thumbnail generation are being added as well.

On the model side, Adobe already offers third-party models from Google, Runway, ElevenLabs, Black Forest Labs, OpenAI, Kling and ByteDance. Its recent acquisition Topaz Labs is releasing two image models: Wonder 4 for upscaling and denoising, and ImageHDR for converting standard images to HDR.

Why it matters

Adobe is betting that agents become the primary interface to creative software rather than a bolt-on feature. The MCP move in particular would make Creative Cloud programmable by whichever assistant a user already prefers — ChatGPT, Claude or Gemini — which keeps Adobe relevant if chatbots become the starting point for creative work. It also puts Adobe on the opposite side of a live debate: TechCrunch notes that Figma co-founder Dylan Field argued last month that MCP is an imperfect standard that benefits frontier labs more than toolmakers. The Verge ties the announcements to a plan Adobe's creativity head David Wadhwani laid out in April, when he said the creative agent would eventually surface "in virtually everything we develop." The caveats are real — the headline MCP capability is months away from even a private beta, and how much agents should be allowed to touch project files is, by Adobe's own admission, still being worked out.