Astron vs
SimpleFunctions.

Based on Astron's public website, Raven 1.0 is Astron's flagship AI agent product for prediction markets. It claims up to 98% short-term forecasting accuracy and offers features including sentiment analysis, tokenized automated strategies, and execution across sports betting and prediction market platforms. Astron's focus is on delivering a packaged product for traders rather than an open API layer or MCP integration for custom agent development.

POST /api/thesis/create takes a plain-language claim — for example, 'The Fed will cut rates before November' — and decomposes it into a causal tree of testable sub-claims. SF then scans Kalshi and Polymarket for contracts that map to each node, propagates probabilities through the tree, and runs a continuous evaluation heartbeat: news scan → price refresh → milestone check → LLM evaluation → confidence update. You can inject external signals via POST /api/thesis/{id}/signal. Public theses are forkable. No competitor publishes a comparable structured decomposition system.

Yes. Portfolio Autopilot uses a 1M-context LLM that reads 13 data sources and passes a 7-gate risk cascade — kill switch, position limits, drawdown gate, regime check, and others — before placing any order on Kalshi or Polymarket. Unlike Astron's tokenized strategy products, SF's Autopilot is a general-purpose autonomous agent that operates on the contracts identified by your thesis, within risk parameters you define per session.

GET /api/calibration returns SF's live Brier scores, broken down by venue, category, and price bucket. Current figures: Kalshi 0.20, Polymarket 0.12 on T-24h price, past 90 days. You can re-run this with a plain curl request at any time. Astron claims up to 98% short-term accuracy for Raven 1.0 but does not publish a Brier score breakdown, a defined time window, or a methodology document against which the claim can be checked.

Based on publicly observable information as of 2026-04, Astron does not publish an MCP server. SimpleFunctions ships a 56-tool MCP server that installs with a single command: claude mcp add simplefunctions --url https://simplefunctions.dev/api/mcp/mcp. It works with Claude Code, Cursor, and any MCP-compatible client, giving agents access to market queries, thesis creation, signal injection, cross-venue pair matching, and trade ideas without writing any integration code.

Astron explicitly covers sports betting markets and its sentiment analysis layer targets that vertical. SimpleFunctions covers Kalshi and Polymarket; it does not currently offer sports betting data or execution. If sports betting coverage is a hard requirement for your use case, Astron is the stronger fit. SF is focused on regulated prediction market venues.

Astron's tokenized strategies are packaged, pre-defined products that traders deploy through Raven 1.0. SF's Portfolio Autopilot is a general autonomous agent — it reads 13 live data sources, runs a causal thesis evaluation cycle, applies a 7-gate risk cascade, and places orders based on your defined thesis and risk parameters. The distinction is product versus programmable agent framework: Astron's model is closer to a fund-in-a-box; SF's is closer to an infrastructure layer you configure and extend.

Yes. The /screen page surfaces computed indicators across 48K+ contracts; trade ideas with conviction and catalyst appear at /api/public/ideas; and the 56-tool MCP server lets Claude or Cursor operate the full system in natural language. Portfolio Autopilot can execute trades autonomously within the risk parameters you set, requiring no manual order entry. For researchers who prefer code, the full REST API and CLI (npm i -g @spfunctions/cli) expose every capability programmatically.