Integrations

Use SPARKIT from the tools you already use.

Claude Desktop and Cursor via the MCP server. Python via the SDK. Anything else via the HTTP API. One install line per integration; cited Markdown reports back from the literature in roughly 90 seconds.

MCP server (Claude Desktop, Cursor, Claude Code)

sparkit-mcp is the SPARKIT Model Context Protocol server. Once configured, your LLM client can hand off literature questions to SPARKIT mid-conversation — no more leaving the chat to curl an API.

Install

uv tool install sparkit-mcp

Or pip install sparkit-mcp. Then grab an API key.

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json on macOS (or %APPDATA%\\Claude\\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "sparkit": {
      "command": "sparkit-mcp",
      "env": { "SPARKIT_API_KEY": "sk_sparkit_..." }
    }
  }
}

Cursor

Edit ~/.cursor/mcp.json with the same block. Reload the window.

{
  "mcpServers": {
    "sparkit": {
      "command": "sparkit-mcp",
      "env": { "SPARKIT_API_KEY": "sk_sparkit_..." }
    }
  }
}

Claude Code

claude mcp add sparkit -e SPARKIT_API_KEY=sk_sparkit_... -- sparkit-mcp

Open source under MIT: github.com/SPARKIT-science/sparkit-mcp. Walkthrough on the launch post.

Python SDK

sparkit-science handles authentication, idempotency, polling, and error mapping. One blocking call returns the final report.

pip install sparkit-science

from sparkit_science import SparkitClient

client = SparkitClient(api_key="sk_sparkit_YOUR_API_KEY")

response = client.research(
    "What is the role of BRCA1 in homologous recombination?"
)
print(response.answer_text)

On PyPI.

HTTP API

Drop into any language. Submit a question, poll a job, get a cited Markdown report.

curl -X POST https://jlsteenwyk--sparkit-api-web.modal.run/v1/research \
  -H "Authorization: Bearer sk_sparkit_..." \
  -H "Content-Type: application/json" \
  -d '{"question": "What is the role of BRCA1 in homologous recombination?"}'

Full reference at /docs.

Which one should I use?

Use…	When
MCP server	You're working in Claude Desktop, Cursor, or Claude Code and want literature lookup mid-conversation.
Python SDK	You're building a backend, pipeline, or agent in Python.
HTTP API	Anything else — TypeScript, Go, R, shell, lab notebooks.

Want a different integration? Reach out at info@sparkit.science.

Get an API key