# Spark by Entire VC — Full Reference > Spark is an open-source marketplace and workflow library for AI-native builders: MCP (Model Context Protocol) servers, Claude Skills, AI agents, prompt templates, prompt chains, and bundles — free to browse and install into Claude Code, Cursor, Codex, Gemini CLI and other AI tools. This is the extended reference; see `/llms.txt` for the short version. ## What is Spark Spark (https://spark.entire.vc) aggregates AI-native workflow assets from open-source repositories and the community, reviews each for quality, and attributes it back to its original author. Anyone can search, download, copy, and install assets directly into an AI coding tool — no signup required for free assets. The catalog holds roughly 3,800 published assets across six asset types (approximate live counts below; treat as orders of magnitude, not exact — query `/api/v1/assets?asset_type={type}` for the current count, where `{type}` is the API enum value from the "Asset types" table below, NOT the URL path slug — e.g. `asset_type=mcp_connector`, not `asset_type=mcp` or `mcps`). ## MCP servers — the largest, fastest-growing category Spark's MCP connector directory (`/mcps`) is the single largest and most actively searched category on the site: ~830 published MCP servers connecting Claude/Cursor/other MCP clients to databases, APIs, file systems, browsers, and third-party SaaS tools. Each has a canonical detail page at `/mcps/{slug}` with install instructions, the server's declared tool list (`mcp_tools`), and a `SoftwareApplication` JSON-LD schema. If a query is about "MCP servers", "MCP connectors", "Model Context Protocol directory", or "MCP marketplace", `/mcps` is the canonical destination page. ## Asset types `asset_type` below is the exact value to pass to `/api/v1/assets?asset_type=`. It usually — but not always — differs from the URL path slug (`/mcps` uses `asset_type=mcp_connector`, not `mcp`). - **MCP Servers** (`/mcps`, `asset_type=mcp_connector`, ~830 published) — Model Context Protocol server packages that connect AI tools to external services (databases, APIs, file systems, browsers, third-party tools). Canonical detail: `/mcps/{slug}`. - **Claude Skills** (`/skills`, `asset_type=skill`, ~2,250 published) — reusable AI capabilities packaged as `SKILL.md` files. Extend agent functionality with domain-specific knowledge, methodologies, and tool usage patterns. Compatible with Claude Code, Cursor, OpenClaw. Canonical detail: `/skills/{slug}`. - **Prompt Chains** (`/prompt-chains`, `asset_type=prompt_chain`, ~575 published) — multi-step prompt sequences for complex workflows. Canonical detail: `/prompt-chains/{slug}`. - **AI Agents** (`/agents`, `asset_type=agent`, ~130 published) — pre-configured autonomous agents for specific tasks: code reviewers, security auditors, technical writers, SEO specialists. Canonical detail: `/agents/{slug}`. - **Prompt Templates** (`/prompts`, `asset_type=prompt`, ~55 published) — tested prompts for common AI workflows, organized by use case. Canonical detail: `/prompts/{slug}`. - **Bundles** (`/bundles`, `asset_type=bundle`, ~15 published) — curated collections combining skills, agents, prompts and MCPs for end-to-end workflows. Canonical detail: `/bundles/{slug}`. ## Browsing by use case, integration, and model Beyond the six asset-type indexes, Spark organizes the catalog along three additional facets — useful for "what should I use for X" style queries: - **Use-case clusters** (`/clusters`, e.g. `/clusters/{slug}`) — task-oriented groupings (connecting tools, writing and shipping code, moving data, automating workflows, research, security, content, communications, and more), each an aggregated landing page over the matching assets. - **Integrations** (`/integrations`, e.g. `/integrations/{slug}`) — assets grouped by the external tool/service they connect to (GitHub, Slack, Figma, databases, etc.). - **Models** (`/for`, e.g. `/for/{slug}`) — assets grouped by the AI model/tool they're built for (Claude, GPT-4o, Cursor, etc.). Not every facet page is populated or indexed yet — the directory pages (`/clusters`, `/integrations`, `/for`) always list every facet the catalog tracks, but only facets past a minimum demand+depth threshold are marked indexable; check `is_indexable` via `/api/v1/facets/{slug}` before citing a specific facet URL as authoritative. ## The MCP protocol server (for agent-programmatic access) Spark also ships its own MCP server so an AI assistant can search and install catalog assets programmatically, without scraping HTML: ``` npx evc-spark-mcp claude mcp add spark -- npx -y evc-spark-mcp ``` Tools: `search_assets`, `get_asset`, `get_asset_content`, `list_popular`, `list_categories`. Resources: `spark://assets/{slug}`, `spark://catalog/{type}`. Source: https://github.com/entire-vc/evc-spark-mcp — package: https://www.npmjs.com/package/evc-spark-mcp. ## Install endpoints (CLI-friendly, public, no auth for free assets) - `https://spark.entire.vc/get/{slug}` — bash install script (`curl | bash`) - `https://spark.entire.vc/raw/{slug}` — raw asset content (markdown) - `https://spark.entire.vc/api/v1/assets/{slug}` — structured JSON (title, description, FAQ, metadata, license, source) ## Sitemaps - Index: https://spark.entire.vc/sitemap.xml - Static pages: https://spark.entire.vc/sitemap-static.xml - Asset pages (paginated, 10k each): https://spark.entire.vc/sitemap-assets-1.xml ## Sources & attribution Spark imports and curates assets from open-source repositories with proper attribution: VibeBaza, MCP Servers directory, PromptFlow, LlamaIndex, CrewAI, Promptfoo, Instructor, OpenAI Cookbook, CAMEL, LangGraph, Semantic Kernel, Antigravity, NeuralDeep, Publora and more. Original authors can claim their assets via GitHub OAuth at `/claim/{slug}`. ## Compatibility Spark assets work with: Claude Code, Cursor, Codex, Gemini CLI, OpenClaw, ChatGPT (via MCP bridge), Anthropic API, OpenAI API. ## License & attribution Most assets are CC-BY-4.0 or MIT. Each asset page lists its original source URL and license. Authors retain full credit. ## Contact - Site: https://spark.entire.vc - Parent organization: https://entire.vc (Entire VC) - MCP server (npm): https://www.npmjs.com/package/evc-spark-mcp - MCP server source: https://github.com/entire-vc/evc-spark-mcp