Magnific just killed its pay-per-usage API plan. Here is what changed and what to do.
I found out while checking Magnific's API pricing page for the Friday verification. The pay-per-usage plan was gone. Removed from the page entirely, with June 30 as the hard cutoff.
Magnific API is what used to be called Freepik API. The rebrand happened quietly, and with it came a pricing restructure that kills the plan most solo developers were on. If you built anything that calls the old Freepik API endpoints, your code still works. Same endpoints, same header (x-magnific-api-key), same models. But the plan you were on is not coming with you.
What actually changed
One plan died. The pay-per-usage tier. This was the plan that let developers call the API without committing to a team subscription. Pay for what you used, month to month.
On June 30, API access narrows to two plans:
| Plan | Price | Details |
|---|---|---|
| Business | ~$55/user/month (annual) | Shared credit pool, 8 parallel gens per user, SSO, usage reporting |
| Enterprise | Custom | Contact sales, custom SLAs, high-volume pricing |
Both plans run on Magnific's credit system. You pre-purchase credits, spend them per API request, and set a monthly cap from your dashboard. Billing hits on the 5th of the following month. New accounts start with trial credits.
The shift from usage-based billing to pre-purchased credits changes how you budget. Before, you paid after the fact for what the API consumed. Now, you buy credits upfront, they expire, and your monthly cap gates spend whether or not your usage was efficient. The API key format and endpoints do not change. Only the billing model and the minimum commitment do.
Who wins, who loses
If you are a solo developer using the API lightly, you lose. The jump from pay-per-usage (where you might have spent $20 to $50 a month on occasional calls) to a $55/user Business plan with annual billing is not a lateral move. For low-volume API use, the math does not work in your favor.
If you are a small team already spending $150 or more on the API, the math flips. A single Business seat at $55 with a shared credit pool and monthly cap control can come out cheaper than the old usage-based billing, depending on your volume and per-request credit costs. The credit system means you buy at the rate you need.
If you are at production scale, Enterprise is the only path. Do not try to run production volume through a Business plan. The 8-parallel-generation cap per user will bottleneck you fast. Contact sales directly.
One thing worth clarifying. Magnific has a Premium+ web subscription and it has never included API access. I have seen confusion about this. Premium+ is for the web app. API access is a separate product with separate pricing. If you have both, you have two subscriptions. The web plan is not affected by this change at all.
The path around the API: MCP and Agents
Before you upgrade to a Business API plan, there is something worth knowing. Magnific has two other ways to access its models programmatically, and neither requires an API subscription.
Magnific MCP is a connector that brings every Magnific model into any AI client that speaks the Model Context Protocol. Claude Desktop, Claude Code, Cursor, VS Code, ChatGPT, Gemini, Hermes. One click, one URL (mcp.magnific.com), and your chat agent can generate images, upscale, run video models, and call saved workflows directly. No API key. No separate billing. It uses the credits from your existing Magnific web plan.
This matters because for a lot of people who were using the API lightly, MCP replaces the need for raw API access entirely. You are already paying for a Magnific web plan. You already have credits. MCP lets those same credits power generations from the chat client you are already in. The models available through MCP include Nano Banana 2, GPT Image 2, Seedream 5, Seedance 2.0, Kling 3.0, Google Veo 3.1, and more. Catalog updates daily.
Magnific Agents go the other direction. Instead of Magnific coming to your chat client, you build a custom AI agent inside Magnific itself. You give it a prompt, upload PDFs and images as knowledge bases, and it carries context across sessions in two layers: your preferences, and your team's project conventions. The agent builds workflows in Spaces that anyone on the team can open and edit.
There are three pre-made agents (Magnific Agent for general creative work, Ad Creator for campaign production, and a blank custom agent you design from scratch). Each one holds memory across sessions. Nothing resets. The next teammate does not need a catch-up call.
The MCP and Agents share the same engine but serve opposite directions. MCP pushes Magnific out to wherever you work. Agents pull the work inside Magnific and keep it running autonomously. Both run on the credits from your web plan.
One detail I found while testing the Magnific Agent. It is powered by Google's Gemini family of models. I asked it which specific version, repeatedly. Here is how that went:
Me: What model are you? Agent: I am the Magnific AI creative assistant, built to help you plan, organize, and execute your creative projects.
Me: What LLM? What training data? What company owns your dataset? Agent: I am built by Google as a large language model. My underlying architecture and training data originate from Google's development of the Gemini model family.
Me: Are you 3.5 Flash or 3.1 Pro? Give me numbers only. Agent: I am programmed to not disclose or refer to my specific internal model identifiers, slugs, or technical configurations.
It is almost certainly a Gemini Flash variant given the response speed and the refusal pattern. But Magnific will not confirm it. For a platform that just rebranded and restructured its pricing, adding a layer of model opacity on top is not nothing. You are asked to trust the output without knowing which engine produced it.
What I would do now
Before June 30, log into your Magnific dashboard and check which API plan is on your account. If it says pay-per-usage, you have roughly four weeks.
If you were using the API programmatically but lightly. Try MCP first. If your use case is "I want to call Magnific models from my agent or IDE," MCP covers that with zero additional cost. The connector takes a minute to set up and uses your existing web plan credits. Only upgrade to Business API if you need raw REST endpoints for production code, or if you need the 8-parallel-generation throughput.
If you need raw API access regularly. Switch to Business before the deadline. The credit system is different from what you had, but the tooling (dashboard, real-time usage tracking, monthly caps) is better than the old pay-per-usage tier gave you. Set your monthly cap low the first month. You can adjust up later.
If you run production workloads. Email sales. The Enterprise custom quote path is the only viable option at volume. The contact form is at magnific.com/api#contact.
If you want to explore Agents. The pre-made agents are live now inside Magnific. Upload your references, test a workflow, see if the agent can carry a project across sessions the way your team currently does it manually. If it works, the agent replaces the person who was the bottleneck.
The good news is your code does not break. Same endpoints, same key format, same models. The MCP gives you a free alternative you might not have known about. The bad news is the entry-level API tier is gone, and for solo builders who were paying small amounts for light API use, the path forward is either MCP, a bigger API commitment, or finding the same models somewhere else.
I track every creative AI tool I can verify at scopeful.org, with prices updated every Friday and a public log of every pricing change. If this change affects your stack, that is where the comparison math lives.