HN Showcase
Back to browse
Polaris – Our API predicts what happens next for AI Agents now with MPP

Polaris – Our API predicts what happens next for AI Agents now with MPP

by JohnWeston·Mar 20, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainBold Bet

Stripe MPP for autonomous agent payments is genuinely novel infrastructure for AI agents.

Strengths
  • •Forecast endpoint includes falsification criteria — rare for prediction APIs to define invalidation conditions.
  • •No API key flow via 402 payment challenges enables truly autonomous agent transactions.
  • •18 verticals with source verification addresses real AI hallucination failure modes.
Weaknesses
  • •Knowledge API space is crowded with Perplexity, Tavily, and Exa already established.
  • •Confidence scoring claims need independent verification — easy to assert, hard to validate.
Category
AI/ML
Target Audience

AI agent developers building verification or research workflows

Similar To

Perplexity API · Tavily · Exa

Post Description

Building Polaris — a structured knowledge API for AI agents.

Instead of returning links, it returns confidence-scored intelligence briefs with bias detection, counter-arguments, entity extraction, and source verification across 18 verticals.

Two things we're most excited about shipping this week:

1. Forecast endpoint — POST /forecast with a topic and get structured predictions: executive summary, ranked predictions with confidence scores and evidence chains, key signals to watch, and falsification criteria ("this prediction is invalidated if..."). Standard depth analyzes 50 briefs, deep analyzes 80. 2. MPP — Agents can now pay for intelligence autonomously via Stripe MPP. No API key, no signup. Agent calls /verify with no credentials → gets a 402 with payment challenges → pays $0.02 → retries → gets the verdict.

Other stuff that makes this different: - Living briefs that evolve as more outlets cover the same story - Bias comparison showing how outlets frame the same event differently - Agent memory (sessions that track what your agent already read) - Watchlists with webhook/email alerts - Source consensus scoring (how much do outlets agree?) - 6 SDKs (Python, TypeScript, LangChain, CrewAI, Vercel AI, MCP)

Try it: curl "https://api.thepolarisreport.com/api/v1/search?q=AI+regulation&api_key=demo" Or test MPP: curl -i -X POST "https://api.thepolarisreport.com/api/v1/verify" \ -H "Content-Type: application/json" \ -d '{"claim":"OpenAI is investing billions in data centers"}'

Free tier: 1,000 credits/month. Live radar dashboard at /radar. Interactive API docs with playgrounds at /docs.

Would love feedback please!

Similar Projects