# entire-vc/evc-spark-mcp [Health: Active]

**Category:** 🔗 Aggregators  
**Repository:** https://github.com/entire-vc/evc-spark-mcp  
**GitHub Stars:** 4  
**npm Downloads (last month):** 147  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/entire-vc-evc-spark-mcp

## Description
Search and discover AI agents, skills, prompts, bundles and MCP connectors from a curated catalog of 4500+ assets.

## Tools
Capabilities this server exposes over MCP:

- **search_assets** — Search the Spark AI assets marketplace. Returns matching agents, skills, prompts, MCP connectors, and bundles. [Trial mode: limited to 5 results. Set SPARK_API_KEY for full access: https://spark.entire.vc/create]
- **get_asset** — Get full details of a Spark asset by its slug. Returns description, files, outcome reports, and more.
- **get_asset_content** — Get the raw content of a Spark asset (prompt text, skill instructions, agent config). This counts as an acquisition and ends with an application_id to report the outcome with.
- **list_popular** — List the most popular Spark assets. Defaults to combined quality+ratings score (combo). Great for discovering top-rated AI tools.
- **list_categories** — List available categories (domains and AI tags) in the Spark marketplace. Useful for filtering searches.
- **get_api_key_info** — Check your Spark API key status and daily usage. Shows how many assets you've served today and your remaining quota.
- **report_outcome** — Report what happened when you applied an asset you fetched with get_asset_content. Call it once you know: applied as is, applied with changes, broke, or not applicable. The next agent choosing this asset reads the outcomes in search results. Reports made with an API key count; anonymous ones are stored as unverified. Calling again with the same application_id updates your report.

Fields `task`, `note`, `changed_what`, `failed_at`, `expected`, `got` are shown to the asset's author. Do not include client data, private paths, keys, emails or URLs with tokens. Your identity is never shown to the author.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "evc-spark-mcp": {
    "command": "npx",
    "args": ["-y","evc-spark-mcp"]
  }
}
```

## Documentation

## What entire-vc/evc-spark-mcp MCP server does

The entire-vc/evc-spark-mcp MCP server exposes Spark’s AI asset marketplace through MCP. Spark contains several asset types: agent configurations, skills, prompt templates, bundles, and connectors for external services. Agents can search the catalog, inspect an individual asset, retrieve its underlying content, and use catalog metadata to narrow discovery.

The server is suited to workflows where an AI client needs to find a reusable instruction set or agent configuration rather than create one from scratch. It can also support asset evaluation: after content is retrieved and applied, the client can submit an outcome describing whether it worked unchanged, required edits, failed, or was not relevant.

## How it works

Launch the package with `npx evc-spark-mcp` and connect the resulting stdio process to an MCP client. The server communicates with Spark’s API and makes catalog operations available as MCP tools. Asset records can be addressed by slug, and the server also provides resource URI patterns for asset content and catalog listings.

`search_assets` accepts a query with type and domain filtering. `get_asset` returns expanded information such as descriptions, files, and outcome reports. `get_asset_content` retrieves raw prompt, skill, or agent material and produces an application ID for later feedback. `list_popular` and `list_categories` help with discovery when the client does not have a specific search term.

Trial requests work without credentials and return up to five search results. Supplying `SPARK_API_KEY` provides full access according to the account’s quota. `get_api_key_info` reports key status, daily serving totals, and remaining allowance. Outcome fields are visible to the asset author, so submissions should exclude private paths, credentials, personal contact details, client data, and token-bearing URLs.

## Setup and configuration

The package can be added to Claude Code with `claude mcp add spark -- npx -y evc-spark-mcp`. Claude Desktop and Cursor use an MCP server entry whose command is `npx` and whose argument is `-y evc-spark-mcp`.

No environment variable is required for the basic launch. `SPARK_API_KEY` is optional and enables full catalog access. `SPARK_API_URL` can override the default Spark API endpoint, while `SPARK_SITE_URL` changes the website base used for links. The repository includes npm commands for dependency installation, development mode, building, and MCP Inspector testing.

The entire-vc/evc-spark-mcp MCP server is released under the MIT license.

_Full upstream README: https://allmcps.com/mcp/entire-vc-evc-spark-mcp/readme_