Build with AI – Polotno skills, MCP, and agent tooling

Most of the cost of adopting an SDK is the time your engineers spend learning it. Polotno ships official tooling for AI coding assistants: skills that teach an agent the API patterns, a live documentation server so answers are never stale, and machine-readable docs. The agent your team already works in can write correct Polotno code on day one.

Key features

Official skills

Two skills, published and maintained by us, that an agent loads when it needs them. polotno-sdk carries API patterns and a capability map, so an agent reaches for the right store method instead of inventing one. polotno-design goes further and produces editable designs, visual proofs, exports, and data-driven variants.

Skills install in one command and work across agents. There is a native plugin for Claude Code and for Codex, and a cross-agent installer that covers Cursor, Copilot, Windsurf, Cline, and others in their own formats.

Live documentation search over MCP

Model knowledge goes stale; a shipped SDK does not stop moving. The Polotno documentation MCP server indexes the current docs and the community forum, so an agent looks up the real answer at the moment it needs it rather than recalling an older API. It runs over stdio and drops into any MCP-compatible editor.

The same search powers the assistant on the docs site, so a developer and their agent are reading the same source.

Machine-readable documentation

Every docs page has a clean markdown twin: append .md to any URL. There is an llms.txt index for agents that want a map, an llms-full.txt that returns the entire documentation set in a single request, and a condensed product brief for assistants evaluating whether Polotno fits a task.

Agent-friendly by design

The tooling works because the product underneath is already schema-first. A Polotno design is JSON with a documented shape, so an agent can generate one, inspect it, change one field, and render it, with every step verifiable. There is no opaque binary format in the middle to guess at.

Example use cases

Faster evaluation – A developer asks their agent to build a proof of concept and gets working Polotno code, not a hallucinated API.
In-house design automation – Generate templates and data-driven variants from a prompt, then render them through the same pipeline as the rest of your app.
Migration work – Point an agent at the schema docs and let it convert templates from a legacy editor.
Support deflection – Your engineers ask the docs directly from their editor instead of opening a ticket.
Internal tooling – Agent-built scripts for bulk edits, brand sweeps, and template QA against a documented store API.

Developer benefits

Official and maintained, not community-scraped and drifting.
Works with Claude Code, Codex, Cursor, Copilot, Windsurf, Cline, and any MCP-compatible editor.
Live docs lookup keeps agents current with the shipped API.
Same JSON schema for humans and agents, so output is inspectable and testable.
Nothing proprietary to adopt: skills are files in your repo.

FAQs

Which AI coding tools are supported?

Claude Code and Codex have native plugins. Cursor, Copilot, Windsurf, Cline, and other agents install through the cross-agent skills CLI. The documentation MCP server works with any MCP-compatible editor.

Does this require a Polotno licence?

No. The skills, the MCP server, and the machine-readable docs are open and free to install. A licence is for using the SDK in production, as usual.

Will an agent write code against an outdated API?

That is the problem the MCP server exists to solve. The SDK skill queries live documentation rather than relying on what the model memorized at training time.