@thesapientcompany/mcp

MCP (Model Context Protocol) server for the Sapient brain-response API. It lets any MCP client — Claude Code, Claude Desktop, Cursor, etc. — submit content to Sapient, read scan results, query your cross-scan intelligence, and read the Sapient knowledge base.

Sapient predicts how a real human brain would respond to a piece of content (video, audio, image, or text) — second by second — and distills that into a score, a grade, and seven lenses.

Try it with no key

You can install with no API key and immediately see a real Sapient readout — call the demo_scan tool for a precomputed sample scan (score, per-second lens timeline, key moments, summary). To scan your own content, add your SAPIENT_API_KEY (mint one at thesapientcompany.com → Settings → API; new accounts include free starter credit) and use run_scan.

Install

Add it to Claude Code with one command:

claude mcp add sapient --env SAPIENT_API_KEY=sk_live_… -- npx -y @thesapientcompany/mcp

Or install with no key to try the demo first:

claude mcp add sapient -- npx -y @thesapientcompany/mcp

Or run the binary directly:

SAPIENT_API_KEY=sk_live_… npx -y @thesapientcompany/mcp

Claude Desktop / generic MCP config

{
  "mcpServers": {
    "sapient": {
      "command": "npx",
      "args": ["-y", "@thesapientcompany/mcp"],
      "env": { "SAPIENT_API_KEY": "sk_live_…" }
    }
  }
}

Remote / HTTP note

This package is a stdio server (it runs locally and talks to the hosted Sapient API at https://www.thesapientcompany.com). Run this stdio server locally with your sk_live_ key. You can point it at a different API host with SAPIENT_BASE_URL.

Configuration

| Env var | Required | Description | | --- | --- | --- | | SAPIENT_API_KEY | yes | Your sk_live_… key (Settings → API at thesapientcompany.com). | | SAPIENT_BASE_URL | no | Override the API base URL (default https://www.thesapientcompany.com). |

Tools

| Tool | What it does | | --- | --- | | demo_scan | No API key needed. Returns a real, precomputed sample scan so you can see a Sapient readout instantly. | | run_scan | Submit content (url or text); optional lenses / options (model defaults to Qualia); set wait: true to poll to completion and return the full result. | | get_scan | Fetch one scan by scan_id — { status } while running, the rich result (timeline, moments, summary) when complete. | | list_lenses | The 4 selectable lenses with one-line descriptions. | | my_history | Your scan history (GET /v1/scans). | | my_intelligence | Your cross-scan patterns + semantic recall (GET /v1/intelligence?q=). | | my_wallet | Your API wallet: balance, free credit, scans remaining (read-only). |

Model — Qualia

Scans run on Qualia, the active model — the visual specialist that reads video (visual + audio). Just omit model (it defaults to Qualia). The mary model is temporarily unavailable.

| Model | Price | Reads | | --- | --- | --- | | qualia (default) | $2.50 / scan pay-as-you-go — volume discounts down to $1.00 / scan at scale | Video (visual + audio) | | mary | — | Temporarily unavailable |

The lenses

The 4 selectable lenses: attention, purchase_intent, manipulation, memory. The default top-3 (when you don't choose) are attention, purchase_intent, manipulation.

Resources

| URI | What it is | | --- | --- | | sapient://knowledge-base | Full-text explanation of how the Sapient model works — the pipeline, the 7 networks, the lenses, and how to read a scan. |

Errors

• 401 → SAPIENT_API_KEY is missing/invalid — set a valid sk_live_ key.
• 402 → out of credits — add API credits in Settings → API.
• 403 → a paid plan is required to use the API.
• 404 → no such scan, or it isn't yours.
• 429 → rate limited; retry with backoff.

Build from source

npm install
npm run build   # tsc → dist/

License

MIT