Magnific adds MCP to paid plans with Slack, auto layers, and resize export

• Magnific turned Magnific MCP live on all paid plans, and magnific's paid-plan update tied that wider rollout to a connector that already works across Claude, ChatGPT, Cursor, Codex, and more.
• magnific's Slack demo showed the most concrete new workflow: Slack, Claude, and Magnific chained together so an assistant can generate assets, extract editable layers, and export resized variants without bouncing between apps.
• The official MCP docs frame the connector as OAuth-based rather than API-key based, while magnific's Gemini reply adds that Gemini support is live too.
• Magnific's own replies clarified the pricing catch: according to magnific's model-selection reply, its unlimited-generations reply, and the official MCP page, MCP actions always consume credits, even on plans that include unlimited generations on the web.

You can browse the MCP page, read the setup docs, and dig through Magnific Academy for tutorials. The weirdly useful detail is that Magnific says the connector can sit beside other tools in the same conversation, so the same chat can pull an image from Magnific and hand it off to Slack or Gmail from there.

The June 10 demo is the cleanest explanation of what changed. In one thread, Magnific showed Slack feeding into Claude and then into Magnific MCP, with three concrete actions:

1. Auto layers, which extracts editable layers.
2. Auto resize, which exports assets into multiple formats.
3. Live adjustment, so edits happen without re-prompting from scratch.

That makes MCP feel less like a prompt pipe and more like a control surface for production art tasks. The official MCP page backs that up with the same editing verbs: upscale, relight, edit layers, and resize.

Magnific launched Agents, MCP, and Flows together at Upscale Conf SF, and magnific's launch post framed the pitch around keeping a human in the loop: direct the system, watch it work in real time, and edit live.

The official MCP landing page is broader than the tweet copy. It lists Claude Desktop, Claude Code, ChatGPT, Cursor, Codex, VS Code, Gemini, and Hermes, and the docs page says setup is just the remote endpoint https://mcp.magnific.com plus an OAuth approval.

The tool surface is also wider than simple image generation:

• Generation across image, video, audio, vector, and 3D, per the official page.
• Editing tools like upscale, relight, layers, and resize, per the official page.
• History search and creation retrieval inside chat, per the docs.
• Model routing in auto mode, with manual model selection when you ask for a specific model, per the official page.

Magnific is pushing two modes at once. One company reply says you can stay in chat for iteration, then jump back into Magnific when you want a more visual place to review outputs, compare versions, manage workflows, or make finer adjustments.

That split shows up in the product docs too. The MCP page says you can describe a workflow, save it as a Space, call it again by name, and share it with a team. Another magnific reply compresses that into a blunt promise: describe a workflow, save it, run it from anywhere.

Early community posts were already testing that side of the launch. BLVCKLIGHTai's first look showed Magnific's new agent and flows with connected nodes, while a follow-up reply called one setup a "beast flow," even if the result needed more tuning.

The most important caveat is pricing behavior. The official MCP page says every paid plan includes MCP access, but every MCP action that generates or transforms content consumes credits.

Magnific repeated that in public replies several ways:

• One reply on unlimited generations says unlimited generations are not available through MCP.
• Another reply on paid plans says even web plans with unlimited generations still spend credits through MCP.
• A model-specific reply says the same rule applies when models like Nano Banana 2 or Kling 2.5 are accessed through MCP.
• A separate reply on cost preview says you can ask the connector for the credit cost before running a request.

One extra useful detail lives only in the docs: according to the official FAQ, checking your balance, browsing old creations, and displaying them in chat do not consume credits. Only actual generation or transformation does.