Useful agents need public evidence from X, YouTube, Reddit, LinkedIn or TikTok. UnifAPI packages that data into Skills, MCP and one live API.

In short: This launch note explains the core problem behind the product: agent workflows need public evidence, but raw vendors and API marketplaces make the integration layer too noisy.

Every interesting agent application we've ever built has eventually needed the same thing: a way for the model to read public evidence. Find tweets about a product. Watch a YouTube transcript. Pull a Reddit thread. Compare public creator profiles.

Doing that today means assembling a bag of vendors. Each one ships a different request shape, a different auth scheme, a different billing model, a different reliability bar. The plumbing eats more engineer-time than the agent itself.

The next version of UnifAPI makes that less abstract. Instead of asking users to start with endpoints, we start with Skills: KOL pricing, creator research, social listening, competitive intelligence. The agent gets a task, then calls the public-data layer through MCP when it needs evidence.

The trade we kept making

Either we picked a hyperscaler scraping platform (Apify, Bright Data) and learned their actor model, or we bought a slice from a specialist (SerpAPI, Firecrawl, Tavily) for one capability and stitched it together with whatever we could find on RapidAPI for the rest.

RapidAPI is the worst of both worlds: thousands of APIs, none of them consistent, billed through subscriptions with hard request caps that punish growth. The moment your usage spikes, you either pay for 10x the tier or you get cut off.

What UnifAPI is

UnifAPI is a Skills-first public-data layer for the subset that agents actually need today. The live catalog includes TikTok, LinkedIn, Instagram, YouTube, Twitter/X, Threads, Reddit, and Hacker News. The API layer is still there; Skills are the front door.

We don't do OAuth-based connectors to your own accounts. That's Composio's territory and they do it well. We're the other half: everyone else's data, on the open web, agent-ready.

What we believe

Pricing should be usage-based. Subscriptions and hard caps belong to a previous era of SaaS billing. We charge $0.001 credits, publish operation minimums from OpenAPI, and keep the bill tied to actual public-data work. We don't punish growth.

DX is the product, but the user experience is now agent-native. Predictable JSON, OpenAPI, and HTTP still matter for builders. MCP and Skills matter for the agent loop: ask, call public-data tools, synthesize, follow up, rerun.

We're early. If you're shipping an agent that reaches for the open web, come talk to us — we're picking a small set of design partners now.

What to read next

Agent First public-data Skills - updates the launch story with the current Skills-first framing.

UnifAPI vs RapidAPI - compares the marketplace problem against UnifAPI's focused surface.

Live API catalog - shows the shipped data layer behind the story.