Adobe plans to let third-party AI agents operate Photoshop, Illustrator, Premiere and After Effects through local Model Context Protocol (MCP) integrations, the company previewed at its MAX 2026 conference in Miami on November 101,6. A private beta is planned for early 20274,5.
The integrations would allow external agents — including ChatGPT, Claude and Google Gemini — to make edits inside native Creative Cloud project files on a user's desktop, with layers, vectors and compositions remaining fully editable afterward. Adobe plans to start with local endpoints and add cloud access later3.
Outside agents, Adobe's rules
Adobe would control which functions, models and algorithms are exposed through the MCP endpoints. Users could also work with Adobe's own Adobe Creative Agent, which is available inside its applications and is expanding to Premiere, Illustrator, InDesign, After Effects and Frame.io13,14.
"Adobe's goal is to help users work the way they want, wherever they are," said Anil Chakravarthy, Adobe's president and incoming chief executive, in a keynote address. Deepa Subramaniam, vice president of product marketing for Creative Cloud, identified Claude, ChatGPT and Gemini as services where Adobe's connector tools let users begin creative tasks before moving into its applications.
Ely Greenfield, Adobe's chief technology officer, framed the move as a response to shifting user behavior. "The separation of creative technology from the creative process is accelerating," Greenfield said at a media briefing on November 107. He added that even if creative work moves to other platforms, Adobe intends to keep its tools accessible wherever users create.
Firefly expansion and new features
Alongside the MCP preview, Adobe expanded its Firefly generative AI platform. Firefly Builder, now in private beta, creates prototypes of websites and interactive apps from natural-language prompts2. A new AI-powered video tool in Firefly assembles videos from clips, images and sound effects9. Adobe also previewed always-on agents that handle recurring tasks in the background, with a beta planned for early 2027, initially in Firefly.
Across its desktop apps, Adobe added AI-driven features: Photoshop gains lighting controls adjustable via sliders; Premiere adds a professional-grade color grading system and object removal from video; Lightroom receives an Open Eyes feature in beta and one-tap blemish removal. Firefly Design, a single-canvas design editing tool, is in beta and included with Creative Cloud.
Firefly is also integrating AI models from partners including Topaz Labs, ByteDance, Google, OpenAI and ElevenLabs. Topaz Labs contributed two image models: one that increases resolution and removes noise, and another that converts standard images to HDR.
Partnerships with Disney, Google and Microsoft
Adobe announced a multiyear strategic partnership with The Walt Disney Company under which Disney plans to develop AI models tailored to its intellectual property using Adobe's Firefly Foundry, with applications in concept exploration and theme park and cruise experience development11. Google plans to put Acrobat in Drive and Gmail, expand Adobe connectors in Gemini and Gemini Enterprise, and explore a near-real-time Photoshop image model with Adobe and Google DeepMind. Through a collaboration with Microsoft, Adobe plans to offer Photoshop, Premiere and After Effects optimized for Arm-based Windows PCs, and Adobe Express creation tools are to be built into the Microsoft Photos app.
CEO Shantanu Narayen said in his keynote that creators face the challenge of producing more content than ever and monetizing it, and that Adobe plans to help them scale content and boost productivity by automating repetitive or complex tasks with AI8.
ANALYSIS By exposing its desktop tools to outside agents while retaining control over which functions are accessible, Adobe is positioning Creative Cloud as infrastructure that third-party AI workflows pass through rather than bypass.