Glif

Glif has evolved into an AI creative agent whose pitch is simplicity: one prompt, every creative AI tool. You describe what you want to make — a launch campaign, an icon set, a short film — and Glif works out which models to use for each step, generates the assets, evaluates them against the brief, regenerates the weak ones and assembles the final output. The site demonstrates this with examples such as turning a single product brief into a lookbook, or producing a researched short video with consistent scenes, narration and music. That agentic loop is the core differentiator versus using individual generative tools one at a time.

Under the hood, Glif is a model-agnostic orchestration layer over image, video, audio and language models, and it is explicit that it adds its own evaluation tooling on top — useful because the hardest part of generative work is often judging which of several outputs is actually better. Context is treated as a first-class asset: add references and characters once, and Glif reuses them across a project so consistency survives many generations. The platform also exposes skills through an MCP server, letting you drive Glif from Claude, Cursor, ChatGPT, Codex or VS Code and keep your context with you.

Who is it for? Marketers, e-commerce teams, educators, musicians and creators who want finished-looking assets without becoming prompt engineers. The trade-offs are those of any agentic, credit-based tool: costs vary by the models used per task, a vague brief still needs iteration, and output speed depends on the underlying engines. But for teams producing a steady stream of visual content, consolidating that work into one chat with persistent context can replace a lot of tool-juggling.