FinFeedAPI vs
SimpleFunctions.

SF's current indexed universe covers Kalshi and Polymarket — 48K+ active contracts. FinFeedAPI adds Manifold and Myriad, which is an advantage for teams that specifically need those venues. SF's value proposition is the agent layer built on top of the venues it covers: thesis decomposition, calibration scoring, computed indicators, MCP tools, and autonomous autopilot — none of which FinFeedAPI ships regardless of venue count.

The thesis system is an agent-layer construct. You POST a natural-language sentence — 'China will devalue the yuan before Q3' — and SF decomposes it into a causal tree of testable sub-claims, maps each branch to live contracts on Kalshi and Polymarket, estimates an implied probability, and schedules an auto-evaluation heartbeat that runs news scans, price refreshes, milestone checks, and LLM evaluations on a recurring cycle. You can inject external signals via /api/thesis/{id}/signal and fork public theses. No prediction market data feed — FinFeedAPI included — ships this layer.

Autopilot is SF's autonomous trading agent. At each tick it loads a 1M-context LLM with 13 data sources — live prices, computed indicators, thesis states, calibration scores, world snapshot, and current portfolio state — then passes the output through a 7-gate risk cascade: kill switch, position limits, drawdown gate, regime classifier, liquidity check, conviction threshold, and concentration cap. Only positions clearing all seven gates reach execution. The LLM reasons over evidence; the gates enforce risk discipline independently.

SF pre-computes six indicators across all 48K+ active contracts: implied yield (IY, annualized return at current price), cliff risk index (CRI, convexity-weighted price distance from a binary resolution), liquidity availability score (LAS, depth-adjusted slippage estimate), event overround (EE, total probability mass across a multi-outcome event), τ-days (calendar days to settlement), and regime label (adverse-selection classification). All are available at /screen and queryable via MCP tools.

Yes. Run claude mcp add simplefunctions --url https://simplefunctions.dev/api/mcp/mcp and SF's 56-tool MCP server is available in your agent context. Tools include market search, thesis creation, orderbook fetch, indicator lookup, world snapshot, cross-venue pair query, and trade ideas. FinFeedAPI does not publish an MCP server, so integration from an agent IDE requires writing HTTP client code manually.

GET /api/calibration returns SF's own Brier scores broken down by venue, category, and price bucket over the past 90 days. Current values: Kalshi 0.20, Polymarket 0.12 on T-24h price. This matters because calibration tells you how much to trust a probability estimate — a market priced at 0.80 that resolves YES only 60% of the time is systematically overconfident. Most prediction market data products do not publish their own accuracy numbers. SF does, and the endpoint is public so you can re-verify with a single curl call.

Based on public information, FinFeedAPI targets enterprise customers requiring SLA-grade data feeds and does not publish a free or self-serve tier. It is designed for production pipelines at scale. If you are a solo developer, researcher, or small team building agents or analytics, SF's public REST API and MCP server require no authentication for read operations and no enterprise contract — you can start querying immediately with no setup cost.