Spotlight: szum — one chart spec, styled everywhere it lands
Most charting tools give you a blank canvas and a hundred config options, and then you spend an afternoon nudging axis labels until the thing looks less like a spreadsheet threw up. Szum takes the opposite bet: it is opinionated on purpose. You write a compact JSON definition, pick one of six curated themes, and the chart comes out looking considered, not because you tuned it, but because you didn't.
The interesting decision here isn't the themes themselves, it's what they're attached to. The same JSON definition renders as an editable Figma asset, an interactive embed you can drop in a product or a Slack message, or an image served through an API. Change the theme once and typography, color, axes, and spacing all move together across every one of those outputs. That's a real problem solved: normally your product charts, your deck charts, and your marketing charts all drift apart because three different people (or three different tools) made them. Here they can't drift, because they're the same spec.
The other bet worth noting is the MCP server. Szum is positioning itself as something an AI agent can call directly, describe the chart in plain language, agent writes the grammar, chart comes out already styled. That's a sensible wedge right now: as more work gets delegated to agents, "can it make a chart that doesn't look like default matplotlib" is a small but real gap, and szum is aiming squarely at it alongside the more obvious QuickChart and Google Image Charts comparisons on their site.
Who should actually use this: teams and solo builders who need charts in more than one place, an app dashboard, a Figma deck, a weekly email, and want them to look like they came from the same design system without hiring someone to make that true. Newsletter writers and indie hackers who currently screenshot a chart from Excel and wince a little are a good fit. So are teams building on top of AI agents who want chart output that doesn't need a designer pass afterward.
Who it's not for: anyone who needs a chart type or interaction outside the curated set. This is not D3, there's no promise of bespoke visualizations, streaming real-time dashboards, or deep customization beyond theme overrides and a custom Google Font. If your job is building a novel visualization nobody's seen before, the opinionated grammar that makes szum fast will feel like a ceiling. And the free tier is real but capped, 5,000 server renders a month, or 250 per IP with no account, so anything render-heavy will push you toward the $29/month Pro plan pretty quickly.
Pricing is straightforward: free covers the editor, Figma plugin, MCP server, and all six themes with a modest render allowance. Pro at $29/month adds brand-level customization (custom themes, any Google Font), removes szum's own branding from published charts, and raises the render ceiling to 100,000 a month with overage. That's a normal SaaS ladder, no surprises, which itself is a small point in its favor.
Where I think this goes: the agent angle is the one to watch. If AI coding assistants and internal tools increasingly need to hand back a chart instead of a table, having a small, well-designed grammar that any agent can target is a genuinely useful default to exist. The risk is the same as the strength, opinionated defaults age well until someone needs the one thing they don't do, and then the whole pitch (don't think about it, it's already right) stops being true for that user. For everyone else, this reads like a sensible, narrow tool that knows exactly what it isn't trying to be.
Try szum: szum.io
See the launch: szum on welaunch.sh