Use Gold Sifter inside the AI workspace your team already uses. Connect a private, read-only MCP endpoint and query the live raise without moving your data into another daily workflow. Tenant-scoped, revocable, and cited against the same records as the product.
Five read-only tools over your live records. Answers come with the same honesty rules as the product: self-reported stays labeled self-reported.
raise_summaryTarget, closed, stated interest, percent to target, tier counts.
pipelineEvery stage with counts, stated value, and the top investors in each.
top_investorsRanked by score, filterable by tier. "Who do I call first?"
find_investorOne person's full record: score, reasoning, stage, source, CRM link.
source_qualityPer-channel leads, A-tier rate, self-reported accredited, stated dollars.
Every engagement gets a private MCP URL, minted per user and revocable any time. It looks like:
https://goldsifter.paperblueprint.com/mcp/gs_your-private-token
Don't have yours? Request MCP access and it arrives with your onboarding.
Individual plans: Settings → Connectors → Add custom connector → paste your endpoint URL. Team and Enterprise organizations: a workspace owner adds the connector at the organization level first. Then just ask: "Who should I call first on the raise?"
ChatGPT Business, Enterprise, and Edu: an authorized workspace administrator enables developer mode, then creates a custom app from Workspace Settings → Apps and adds your endpoint as a remote MCP server. Availability and controls vary by plan and workspace role.
Add to your MCP config:
{
"mcpServers": {
"gold-sifter": { "url": "https://goldsifter.paperblueprint.com/mcp/gs_your-private-token" }
}
}
| Transport | Streamable HTTP (JSON-RPC), MCP protocol version 2025-06-18 |
| Scope | Read-only. Five tools over your tenant's records; no write, send, or configuration surface exists on this endpoint |
| Token | Private per user, issued by the operator at onboarding. Long-lived until revoked; revocation by email is immediate, and rotation is available on request |
| Isolation | Tenant-scoped: a token can only ever read its own engagement's records |
| Logging | Endpoint access is logged for security review |
| Rate limits | Fair-use; no hard limit published. Sustained abuse revokes the token |
| Data path | Answers are computed from your live records and returned to your AI client. Your AI provider receives those answers, and its own retention and training terms apply to your conversation; check the provider's data controls |
Client setup menus change; the steps above are current as of July 5, 2026. If a menu moved, email for the current path.