Official SDKs

ScienceToStartup is an Agent Operating System for Research Commercialization

API and MCP Platform for Turning Research Papers into Buildable Product Signals.

Use the official client surfaces when you want a stable wrapper around paper discovery, Signal Canvas, remote MCP, and workspace execution without rebuilding auth, endpoints, and action payloads by hand.

Create Key Quickstart API Docs

Use one client surface for REST discovery, MCP calls, and workspace-native execution.

Start from proof objects like papers, topics, and benchmark surfaces, then keep those IDs stable through follow-on actions.

Treat SDKs as ergonomics wrappers, not as alternate contracts. OpenAPI, capabilities.json, and MCP remain the canonical interfaces.

Install (JavaScript / TypeScript)

pnpm add @sciencetostartup/sdk
# or
npm install @sciencetostartup/sdk

Install (Python)

pip install sciencetostartup-sdk

Quickstart (JavaScript / TypeScript)

import { ScienceToStartupClient } from "@sciencetostartup/sdk";

const client = new ScienceToStartupClient({
  apiKey: process.env.STS_API_KEY!,
});

const papers = await client.searchPapers({
  query: "voice agents",
  limit: 5,
});

const workspace = await client.createWorkspaceFromSeed({
  title: "Voice agent shortlist",
  seedType: "paper",
  seedId: papers.items[0]?.arxiv_id ?? "",
});

await client.runWorkspaceAction({
  workspaceId: workspace.workspace_id,
  runKind: "launch_pack",
});

Quickstart (Python)

from sciencetostartup_sdk import ScienceToStartupClient

client = ScienceToStartupClient(api_key="s2s_YOUR_KEY")

papers = client.search_papers(query="voice agents", limit=5)
workspace = client.create_workspace_from_seed(
    title="Voice agent shortlist",
    seed_type="paper",
    seed_id=papers["items"][0]["arxiv_id"],
)
client.run_workspace_action(
    workspace_id=workspace["workspace_id"],
    run_kind="launch_pack",
)

MCP via SDK

import { ScienceToStartupClient } from "@sciencetostartup/sdk";

const client = new ScienceToStartupClient({
  apiKey: process.env.STS_API_KEY!,
});

const tools = await client.listMcpTools();
const result = await client.callMcpTool("search_papers", {
  query: "voice agents",
  limit: 5,
});