Claude Plugin

The official Claude Code plugin for Vovk.ts ships topic-based skills that teach the coding agent how to use the framework when you describe what you want to build. Skills load only when relevant — typing “scaffold a new tenant” pulls in the multitenant skill, “stream chat tokens” pulls in JSON Lines, and so on.

The plugin lives in the same monorepo as Vovk.ts itself, at packages/claude-plugin, and ships alongside the framework on every release.

Why Vovk.ts is built for AI-assisted coding

Vovk.ts’s structure is the prompt — the AI mind model is built into the framework:

• Logic groups under src/modules/<name>/ — Controller + Service per feature, not scattered across lib/.
• Controller / Service split — Controllers define procedures (decorated). Services hold business logic (plain classes). Clean separation between what the endpoint is and what it does.
• Methods on a service class, not loose helpers — fewer files, predictable layout. The assistant finds the right file on the first try.
• Single source of truth — same procedure().handle() powers the HTTP endpoint, the SSR call (.fn()), and the AI tool (deriveTools). No duplication. Less for the model to reconcile.
• Plain REST under the hood — curl works. fetch works. Types flow end-to-end without locking you into a custom protocol.
• Multitenancy baked in — multitenant() proxy + segment-per-tenant; one Next.js app hosts many tenants on subdomains without re-architecting.
• OpenAPI native, AI tools native — schema generated from procedures, Scalar docs auto-mounted, every procedure can become an LLM tool with one line.

Why use the plugin?

Without it, asking Claude to “scaffold a Vovk procedure with Zod validation” tends to go one of two ways: the model hallucinates (training data is months stale and Vovk’s API has moved on), or it fetches vovk.dev mid-task — slow, hits rate limits, and the wrong page often loads first.

With the plugin, the framework’s idioms are loaded as topic-scoped skills the agent already understands:

• Topic-scoped, not all-or-nothing. Skills load only when relevant — “stream chat tokens” pulls in JSON Lines; “scaffold a tenant” pulls in multitenant. Claude doesn’t drown in thousands of lines of docs to answer a focused question.
• Self-contained. A directive at the top of every skill tells the agent don’t fetch vovk.dev mid-task — the plugin is the source of truth. Works offline; predictable cost; no rate-limit surprises.
• Cross-skill handoffs. Skills know to escalate. The mixins skill points at the tools skill for LLM exposure; the procedure skill points at jsonlines for streaming. The agent loads the right context, not adjacent context.
• Caveman-optimized prose — skill markdown is token-tightened (~10% fewer tokens per load, no loss of substance).

Install

/plugin marketplace add finom/vovk

/plugin install vovk@vovk

/reload-plugins

/plugin

Skills

The plugin ships fourteen topic-based skills covering every layer of Vovk.ts:

• vovk:init — initialize Vovk.ts in a Next.js App Router project, or scaffold a fresh Next.js app and run vovk init on top.
• vovk:common — framework fundamentals: packages, vovk.config.mjs, type inference helpers (VovkBody, VovkOutput, …), short API reference.
• vovk:segment — segments (root, named, static), initSegment, segment priority, generateStaticParams.
• vovk:multitenant — multi-tenant routing via subdomains: multitenant() proxy, overrides shape, per-tenant segments and frontend pages, wildcard DNS.
• vovk:procedure — procedures, validation (Zod / Valibot / ArkType), controllers, HTTP decorators, req.vovk, error handling, content types, .fn() for SSR / server components / server actions.
• vovk:decorators — built-in and custom decorators (createDecorator), authorization patterns, req.vovk.meta(), stacking order, decorate() for projects without experimentalDecorators.
• vovk:rpc — generated vovk-client, composed vs segmented clients, call shape, customFetcher, error rethrow, type inference from client methods.
• vovk:jsonlines — JSON Lines streaming: generator handlers, JSONLinesResponder, progressive(), client async iteration, using, asPromise, abort.
• vovk:openapi — OpenAPI 3.x generation: @operation metadata, outputConfig.openAPIObject, per-segment overrides, Scalar docs, _schema_ endpoint.
• vovk:mixins — import third-party OpenAPI 3.x schemas as typed client modules, call them identically to native RPC modules.
• vovk:tools — expose procedures as LLM tools via deriveTools(), MCP-compatible output, @operation, controllers vs RPC modules, OpenAI / Anthropic / MCP wiring.
• vovk:bundle — vovk bundle CLI for publishable TypeScript SDKs.
• vovk:python — generate a typed Python client (vovk-python), py / pySrc templates, TypedDict shapes, JSON Lines via Python generators, PyPI publishing.
• vovk:rust — generate a typed Rust crate (vovk-rust), rs / rsSrc templates, async reqwest call shape, futures::Stream consumption, crates.io publishing.

First prompts to try

The skills trigger automatically when you describe what you want to build. Pick the type of project you’re starting:

• Greenfield — “Set up Vovk.ts in a new Next.js project. I want a /api/tasks CRUD endpoint with Zod validation, and a Next.js page that consumes it through the typed client.”
• Existing Next.js project — “Add Vovk.ts to my existing Next.js app and scaffold a UserController with getUser / createUser.”
• Stream-heavy work — “Add a /api/chat JSON Lines streaming endpoint that proxies OpenAI completions, plus a Python script that consumes the stream.”

Reporting bugs

If a skill produces wrong code or contradicts itself, that’s a plugin bug — open an issue at github.com/finom/vovk/issues with the prompt you used and the skill that loaded.

More info:

• Plugin source 
• Claude Code docs — Plugins