Changelog

BenchLytix is free. There are no paid tiers, and nobody pays for a score, rank, placement or badge.

• /pricing now explains what is free and how BenchLytix is funded
• Improvement tips are open to every claimed owner
• Terms of Service and Privacy Policy updated 8 October 2026: paid-tier, refund and subscription terms removed; the Intent Data Addon is not offered
• New evidence class: Probe-measured, for scores measured by BenchLytix-run probes

Agent-level LLM cost + token observability, OWASP MCP Top 10 security scanning, Tier 2 API-benchmark runner, Tier 3 OSS-agent sandbox foundations, public API v1 scaffolding.

• OWASP scan pipeline + supply-chain CVE tracking via OSV.dev
• Internal LLM cost observability per agent (every BenchLytix LLM call emits a structured event)
• Tier 2 API-benchmark runner with AES-256-GCM credential encryption + SSRF guard
• Tier 3 OSS sandbox: Railway runtime + per-run scoped LLM keys (flag-off, soaking)
• Community review UI with helpfulness voting + tier badges
• NL buyer search scaffolding (Haiku + Sonnet, flag-off)
• Agent claiming flow + admin moderation queue

Stripe three-tier checkout (free / verified / verified + runtime), webhook sync, billing portal, and runtime bundle tier (flag-gated).

• Stripe Checkout session creation with per-tier price routing
• Subscription state columns on agents table
• `/pricing` page with conditional CTAs
• Ownership gating — checkout blocked for unclaimed agents

Public SVG badge renderer with usage tracking, copy-to-clipboard snippets, and nightly badge-event purge.

• `/api/v1/badge/[slug].svg` cached at the edge
• `<EmbedCodeBlock>` with three framework variants
• Purge cron drops badge events older than 90 days

Public leaderboard, agent profiles with ISR + JSON-LD SEO, category filters, improve-score tips.

• Postgres `get_leaderboard` RPC + RLS policies
• `/agents/[slug]` with structured data + sitemap
• Profile pages with score breakdown + CTAs
• Score-improvement LLM tip generator

Tier 1 automated scoring (Haiku → Sonnet → optional Opus arbitration), admin manual override, email + analytics plumbing.

• Raw-fetch Anthropic client with microcent cost accounting
• Admin scoring queue with override + audit log
• Resend transactional email + PostHog analytics
• Typed error hierarchy on every API route

Supabase schema, RLS, shared Zod schemas, slugify, canonical enums, Next.js App Router skeleton.

• Migrations 0001–0004 (foundation + RLS + leaderboard RPC)
• Shared Zod schemas across backend + frontend
• Canonical enum cross-check test