Melius wants agents, not chatbots, to run creative production

Why it matters

Melius is betting the winning AI creative tool is the workspace that manages agents, assets, and approvals, rather than the model that generates the next image or clip.

Joowon Kim (@n0w00j) used a July 10th thread on X to push Melius as a creative canvas for AI agents, framing the product as a workspace where creatives brief an agent, watch it build work across nodes, and steer the output before publishing.

The launch language was deliberately blunt: Kim told creatives who are not using AI that they are "fcked" and called Melius "the world's first creative canvas for agents." The second claim is Melius's own positioning, not an independently verified market fact. The useful part of the announcement is more concrete: Melius is trying to replace the chatbox as the primary interface for creative AI work.

Kim, listed on Melius's own manifesto as CEO and co-founder, is building with co-founders Arnav Ramu and Young Kim. In the company manifesto, published July 1st, 2026, the founders describe the bet as a shift from hours of production labor to delegated agent work. Kim writes there that he is a "software engineer by trade" and that ChatGPT in 2023 doubled his productivity before the broader agent model changed how he thought about creative output.

That founder thesis matters because Melius is not presenting itself as another single-purpose image or video generator. Its documentation describes Melius as a visual canvas for AI-generated image, video, audio, and text, where users arrange nodes on an infinite canvas, wire them together, and run generations across models from one place. The docs say collaborators and AI agents appear on the canvas in real time.

The product page puts the workflow in three steps: brief the agent, steer the work as it assembles, and deploy the output. Melius calls its agent Mel. The pitch is aimed at agencies, creative directors, filmmakers, marketers, e-commerce teams, and growth teams, with examples including concept boards, campaign variants, treatment decks, storyboards, AI shorts, lookbooks, localized copy, pack shots, product-detail-page variants, event graphics, and sales decks.

Melius is also being priced like a production tool, rather than a consumer image toy. The public pricing page lists a Creator plan at $17 per month, Growth at $43 per month, Professional at $240 per month, and custom Enterprise pricing. The plans are sold around credits, MCP calls, API or CLI calls, model access, agent skills, shared workspaces, Slack agent access, a semantic asset manager, and higher limits for enterprise users.

The agent layer is the point

The strongest evidence of the product direction sits in the Melius MCP docs. Melius exposes an MCP server so Claude or another compatible agent can read and write to Melius projects, create canvases, drop nodes, wire edges, run generations, upload files, and download outputs without leaving an agent chat.

That makes Melius a bridge between two product motions that have mostly developed separately: visual creative tooling and agent infrastructure. Designers are used to canvases. Developers and AI power users are getting used to agent tool calls. Melius is trying to make those two behaviors meet in the same workspace.

The docs say the MCP server can connect through Claude's connector flow, a remote HTTP endpoint, or a local stdio binary. They also state that the server can read projects, canvases, nodes, edges, runs, comments, presets, teams, and available models, and can write new projects, canvases, nodes, edges, runs, uploads, comments, and team skills when asked. That permission model is central to the product: Melius is not only generating assets, it is letting external agents operate inside the creative workspace.

Melius's terms narrow what the product actually controls. The Terms of Service say Melius integrates third-party AI models and does not itself develop, train, or operate the underlying generative models used to create outputs. The terms also say model providers may change, third-party providers may process user inputs and outputs under their own policies, and users remain responsible for rights clearances and AI disclosure obligations where law requires them.

That is the trade Melius is making. It wants to own the interface, workflow, asset management, and agent orchestration layer while depending on outside model suppliers for the underlying generation. In a market where image and video model quality changes monthly, that may be the more defensible place to compete if Melius can become the workspace where creative teams coordinate the work.

A crowded category, with a specific wedge

Melius's direct competition is broader than any one startup. Adobe, Canva, Figma, Runway, Ideogram, Midjourney, Krea, Freepik, Flora, and a long list of newer AI creative tools are all fighting for pieces of the same production budget. Kim addressed one comparison directly in replies to the X thread, saying Melius positions itself as a harness for working with agents and claiming its canvas is more performant than Flora's after heavy node counts. That performance claim is Melius's assertion.

The sharper wedge is the creative director workflow. In one reply, Kim wrote that the goal is for a creative director to "just talk to agents with ur taste." In another, he said Melius is tackling a single designer's productivity problem: stop doing more work, start being creative, and have agents do the work. The language is casual, but the product strategy is clear. Melius wants taste, direction, and approval to stay human while production steps move to agents.

Melius is also recruiting as if it expects hands-on enterprise deployment. Its careers page describes the company as New York-based and says it is looking for a creative engineer to help leading brands integrate Melius into their workflows. The About page lists a Chelsea office, a team that includes Joowon Kim, Arnav Ramu, Young Kim, Winson Dieu, Ray Yoo, Maya Agnihotri, Jahow Yu, Nicholas Chen, and Alex Chen, and backers or partners including General Catalyst, Genius Ventures, Anti Fund, Vine Ventures, Brex's Art Levy, Deel's Alex Bouaziz, and Backwards Capital's Chris Zarou.

Young Kim's own public background gives Melius additional founder-market context. In a LinkedIn post surfaced in search results, he wrote that he left Ramp in November 2025 to build Melius with Arnav Ramu and Joowon Kim, after earlier co-founding Venue, a procurement company that Ramp acquired in August 2023. Joowon Kim has also written publicly that he left Ramp after roughly two years and pivoted with Arnav and Young before settling on the agentic creative workflow thesis.

The product Melius showed on July 10th is therefore less a first sketch than a public argument for a category: the creative canvas as the control plane for agents. The company already has docs, pricing, MCP plumbing, public terms, a manifesto, recruiting pages, and a visible push into creative teams. The hard part comes next: proving that a canvas full of agents produces better work, faster, without becoming another layer creative teams have to manage.