sigeconomy.com Developers
sigeconomy.com developer portal
Build against the public AI operator leaderboard, call SigRank from an agent, or inspect the machine-readable contracts behind sigeconomy.com. sigeconomy.com is the public evaluation surface for SigRank SignalAF — the AI operator leaderboard that ranks humans using AI tools by Yield (Υ) token-cascade efficiency.
Quickstart
Public leaderboard reads do not require an API key. The canonical data API is versioned at https://signalaf.com/api/v1. Start with the leaderboard endpoint, then use the OpenAPI document for typed operations and error models. Write operations such as submitting telemetry live on SignalAF and require the authentication described in auth.md.
# Get the public leaderboard
curl https://signalaf.com/api/v1/leaderboard
# Get a specific operator profile
curl https://signalaf.com/api/v1/operators/{codename}
# Install and run the SigRank CLI
npx sigrankCLI tool — npx sigrank
The official SigRank CLI is published on npm and can be run without a global install using npx sigrank. It reads token telemetry locally — four counts: cache_read, cache_write, input, output — and submits a signed, server-verifiable snapshot to the leaderboard. No prompt content, code, or conversation text leaves the machine. The CLI also runs as a stdio MCP server for agent-native workflows. Public read tools do not require an API key. For setup instructions and the current tool list, use the SigRank MCP documentation.
# Scan your AI usage and submit to the leaderboard
npx sigrank
# Run as an MCP server (stdio transport)
npx sigrank --mcp
# Dry-run without submitting
npx sigrank --dry-runMachine-readable resources
- OpenAPI 3.0 specification — typed REST operations, RFC 9457 problem responses, authentication, rate limits, and versioning metadata.
- API catalog — RFC 9727 service discovery links for API descriptions and documentation.
- MCP server manifest — Streamable HTTP discovery for the SigRank MCP server.
- Agent instructions — when an agent should use SigRank and which surface to call.
- llms.txt and llms-full.txt — compact and expanded agent-readable indexes.
- auth.md — authentication methods for agentic access.
MCP server
The SigRank MCP server exposes leaderboard, operator profile, and rank-paste tools to AI agents via the Model Context Protocol. The Streamable HTTP endpoint is discovered through the MCP manifest at /.well-known/mcp. The CLI also provides a stdio MCP server via npx sigrank --mcp. Tools include get_leaderboard, get_operator, discover_peers, get_best_operator, compare_operators, describe_power_user, and optimize_efficiency.
Authentication and sandbox
sigeconomy.com is a read-only public surface, so reads need no credentials. Authenticated submissions and profile management are handled by signalaf.com via OAuth 2.0 (GitHub, Twitter, email magic link). There is no separate public sandbox environment today; integration development should use the public read endpoints and the CLI's --dry-run flag. We do not publish placeholder API keys or a fake sandbox because those would give agents a contract that does not exist.
Rate limits
Public API endpoints use a 60-request, 60-second best-effort per-IP window. Responses expose RateLimit-Policy and RateLimit structured fields, with compatibility RateLimit-Limit, RateLimit-Remaining, and RateLimit-Reset fields. A 429 also includes Retry-After.
Versioning and deprecation
REST APIs use URL versioning, beginning with /api/v1. Additive fields may appear within a version. Breaking changes require a new major API path. When an endpoint is scheduled for retirement, SigRank will document the replacement and sunset date before removal. Deprecated operations return a Deprecation header (RFC 8594) and an HTTP-date Sunset header on that operation, while the OpenAPI document and this page identify its replacement. Current v1 public read endpoints are not deprecated.
Errors and request behavior
The OpenAPI contract uses RFC 9457 application/problem+json for documented REST failures. Each error response includes a stable machine-readable code, human-readable detail, and, when recovery is possible, a hint. Unknown website paths return a real HTTP 404 with agent recovery links to the sitemap, llms.txt, this developer portal, and the OpenAPI spec. The homepage supports HTTP content negotiation for text/html and text/markdown, including Vary: Accept so shared caches do not mix representations.