Krea’s AI agent doesn’t require an invite anymore. A changelog update on September 10 opened it to every plan, Free included, no access code needed.
What changed
Krea Agent works by chat. You write a brief describing what you want, and the agent plans the job, picks a model from Krea’s catalog of more than 150 image, video, audio, and enhancement models, generates the assets, checks and edits the results, and files everything in a context library your team can reuse. It’s reachable from krea.ai/agent, from Krea’s own product switcher, from Slack, and from a ChatGPT plugin. It also ships as an MCP server, so it can plug into an existing Claude or Cursor session instead of requiring a separate browser tab.
Paid plans layer on more: a choice of orchestrating model (GPT, Claude, or Gemini families, on Business and Enterprise), higher “effort” tiers, model pinning, and, on Enterprise, saved routines and custom agents.
Why it matters for designers
If you already use Krea, the agent piece is no longer gated behind an invite, so it’s worth an actual look, especially the MCP hook. That’s the part with real workflow implications: an agent already living in your Claude or Cursor session can hand off an image, video, or audio generation job without you tabbing over to a browser and picking a model by hand. For a studio weighing whether to add another point tool, this is a low-cost way to test whether agent-orchestrated generation actually saves time over doing it yourself.
What to watch out for
Krea hasn’t published usage-rights specifics for the agent’s output across its full 150-plus-model catalog, and it’s unclear what happens to files placed in the shared context library once a team starts relying on it, so this isn’t somewhere to drop real client assets without checking Krea’s terms first. The September 10 to 15 bonus-credit window has also already passed. Generations now draw on your plan’s regular compute credit balance, not a free allotment.
