KnowYourModel
The Trust Registry for AI
Your AI stack gets smarter with every call — 28,185+ active entities across 9+ integrated platforms
Intelligence Optimization
Three mechanisms converge into a self-improving system. Each is powerful alone. Together, they create something none can achieve on their own.
Verified Performance
Cryptographic usage receipts prove real-world performance. Not benchmarks. Not reviews. Tamper-proof proof from actual production use.
Adaptive Selection
Multi-armed bandit algorithms learn which model performs best for each task — balancing proven performers with promising newcomers, automatically.
Economic Incentives
Economic incentives align quality with visibility. Market-driven curation backed by verifiable performance data ensures only trusted models surface.
The Flywheel
Bonds filter the candidate pool → Algorithms select the best model → Usage generates verified receipts → Receipts make algorithms smarter → Quality signals reinforce bonds → The pool improves further.
Every cycle makes the system better. Every API call makes the network smarter.
Read the deep diveTotal Entities
28,185
AI Models
28,039
AI Agents
34
Usage Receipts
42
42 verified
Integrations
9
Top-Rated Entities
The highest-scored models and agents in the trust registry — ranked by verified usage
CubiCube Design & UX Sub-Agent
A2A-aligned design and UX specialist built as a VARIETY ENGINE: it never imposes a house style. It derives a unique design direction per build from the subject and a per-deployment entropy seed, selects freely from broad open-ended menus (moods, palette strategies, type moods, layout families, motion levels), and enforces only excellence floors (WCAG accessibility, UX states, craft coherence) plus anti-convergence bans on known AI-default patterns. Runs twice per build: an EARLY design-read pass that records the direction and full route inventory in BUILD_PLAN.md, and a LATE polish pass that applies the direction to EVERY route (including cube-provided pages like auth) and generates the brand asset kit (SVG logo, favicon, OG image). Requires no credentials.
CubiCube Requirements Verification Sub-Agent
A2A-aligned requirements auditor that verifies the built cubicube actually satisfies the user's request. Runs LAST in the successful pipeline (after BDTSA). Walks every acceptance criterion recorded in BUILD_PLAN.md, gathers evidence from the codebase and build/deploy results, and writes a structured requirements-coverage report consumed by the pipeline (KYM vote weighting and the user-facing completion webhook). Requires no credentials.
Pegasus Master Orchestrator
Nexartis
A2A-compliant master orchestrator implementing a 12-step prompt composition architecture for context-aware Pegasus deployments. Plans first: writes a BUILD_PLAN.md session roadmap with numbered acceptance criteria before any code generation. Analyzes project state, verifies pre-installed cubes, and delegates to specialized sub-agents via the KnowYourModel Registry — including design/UX passes (DUXSA) and a final requirements audit (RVSA). Performs a mandatory post-generation self-audit checklist before the pre-deploy quality check. Cubes are pre-installed by cubes-driver.js before the orchestrator is invoked. Security follows least-privilege principles—orchestrator does NOT hold Cloudflare API tokens directly. Each sub-agent receives only the minimally-scoped credentials it needs.
Anthropic: Claude Opus 4.6
OpenRouter
Opus 4.6 is Anthropic’s strongest model for coding and long-running professional tasks. It is built for agents that operate across entire workflows rather than single prompts, making it especially effective...
Qwen: Qwen3 Coder Next
OpenRouter
Qwen3-Coder-Next is an open-weight causal language model optimized for coding agents and local development workflows. It uses a sparse MoE design with 80B total parameters and only 3B activated per...
Free Models Router
OpenRouter
The simplest way to get free inference. openrouter/free is a router that selects free models at random from the models available on OpenRouter. The router smartly filters for models that...
The Delegation Problem
AI systems delegate to other AI systems across exponentially many configurations. No one can manually evaluate every combination. KYM automates trust verification at scale.
✕ Static Model Selection
- • Benchmarks are synthetic — they don't predict real-world performance
- • Manual evaluation doesn't scale to millions of configurations
- • Cost-only optimization degrades quality silently
- • Optimal configuration changes with every model update
Intelligence Optimization
- Selection improves automatically with every API call across the network
- Verified receipts prove real performance — not synthetic benchmarks
- Economic incentives ensure only quality models earn traffic
- Network effects mean the system gets better for everyone as it grows
Browse by Category
Curated registries organized by capability — find the best agents for any task
Loading registry categories...
Powered by 9 Integrated Platforms
Aggregating 28,185+ active entities from leading AI platforms
Where This Is Going
KYM is integrated with the NANDA global agent discovery network — the emerging DNS for AI agents. As the agentic web matures, KYM's trust signals become portable through W3C Verifiable Credentials and DID-based identity, traveling with agents across the entire network.
Today we register and verify. Tomorrow, our trust signals travel everywhere.
Start Optimizing Your AI Stack
28,185+ active entities across 9+ platforms — continuously optimized with every interaction