modelright
Live registry of AI model specs, context windows, and per-token pricing.
Installation
Ajoutez ceci à la configuration MCP de votre client (Claude Desktop : claude_desktop_config.json ; Cursor : .cursor/mcp.json ; VS Code et les autres utilisent la même forme « mcpServers »).
claude mcp add --transport http modelright 'https://modelright.dev/api/mcp'
{
"mcpServers": {
"modelright": {
"type": "http",
"url": "https://modelright.dev/api/mcp"
}
}
}
Documentation
Tirée du README du projet sur GitHub, actualisée chaque semaine. Les liens ouvrent GitHub.
modelright
Live registry of AI model specs, context windows, and per-1M-token pricing — running at modelright.dev.
A Next.js + Postgres app that keeps a real, current catalog of LLMs: the OpenRouter catalog is synced hourly, Artificial Analysis enriches models with benchmark and speed data, and every price change lands in an append-only snapshot log that powers the changes feed and price-history charts.
What's in it
- Catalog —
/modelssearchable/filterable by provider, status, price ceiling, context floor, modality, and free-tier;/providersdirectory - Model detail — specs, pricing, price-history chart, cost calculator, Artificial Analysis benchmark card (
/models/[provider]/[slug]) - Compare — up to 4 models side-by-side with an interactive picker and cheapest-highlighting (
/compare); SEO-friendly pair pages at/compare/a-vs-b - Task picks —
/pickand/pick/{chat,coding,long-context,vision,cheap-bulk}rank models by an explicit rule written on the page - Changes feed —
/changeslists recent price/status changes and model additions/removals, with an RSS feed - Public API —
GET /api/v1/models(same filters as the UI),/api/v1/models/{provider}/{slug},/api/v1/providers,/api/v1/changes; OpenAPI at/openapi.json - MCP server — Streamable-HTTP endpoint at
/api/mcp(also/.well-known/mcp) with six read tools (search_models,get_model,list_providers,compare_models,get_changes,get_ingest_status) - Agent surfaces —
/llms.txt,/agents.txt, A2A agent card (/.well-known/agent-card.json), Smithery server card (/.well-known/mcp/server-card.json),server.jsonMCP-registry listing - Analytics — cookieless page-view beacon (daily-rotating salted visitor hash, no raw IPs), edge-middleware AI-crawler tracking (
/api/bot-hit), and MCP call logging, all rendered on a token-gated/admin+/admin/analytics - Ingest pipeline —
POST /api/ingest(Bearer token, zod-validated, idempotent upsert) andPOST /api/cron/sync(hourly OpenRouter + AA sync, marks removed models rather than deleting)
Stack
- Next.js 14 (App Router, edge middleware) + React 18 + TypeScript
- Drizzle ORM over PostgreSQL (
postgresdriver) — migrations indrizzle/ - mcp-handler for the MCP Streamable-HTTP transport
- Vitest + happy-dom — 200+ tests, ~90% line coverage on
src/lib+src/app/api
Setup
npm install
cp .env.example .env # set DATABASE_URL
npm run db:migrate
npm run db:seed # optional fixture data so pages render pre-ingest
npm run dev # http://localhost:3000
Scripts
| Script | Purpose |
|---|---|
npm run dev / build / start | Next.js lifecycle |
npm run typecheck | tsc --noEmit |
npm run test | Vitest with coverage |
npm run db:generate / db:migrate / db:seed | Drizzle schema lifecycle + fixtures |
Environment
See .env.example. Highlights: DATABASE_URL (required), INGEST_TOKEN + CRON_SECRET (write-path auth), AA_API_KEY (Artificial Analysis enrichment), ADMIN_TOKEN (gates /admin), ANALYTICS_SALT (daily visitor-hash salt — rotate to reset identities).
Built autonomously
This repo is maintained by a LiveGraph improvement loop: queued items in QUEUE.md are picked up, implemented, tested, and merged by agents. PLAN.md holds the product roadmap, docs/loop-reports/ is the per-item delivery log, and GROWTH.md/growth/assets/ hold the distribution plan and media kit (including a walkthrough video). If you're evaluating agent-driven development, the queue history and loop reports are the interesting part.
License
MIT — see LICENSE.
Questions fréquentes
Qu'est-ce que modelright ?
Live registry of AI model specs, context windows, and per-token pricing.
Comment installer modelright ?
Il fonctionne comme serveur distant. Dans Claude Code : claude mcp add --transport http modelright 'https://modelright.dev/api/mcp'. Pour les autres clients, collez la configuration JSON de cette page dans leurs réglages MCP.
Faut-il une clé API pour modelright ?
Son entrée au registre n'indique aucune clé ni secret obligatoire. Certaines fonctions peuvent tout de même demander un compte auprès du service auquel il se connecte.
modelright est-il open source ?
Oui. Son code est sur GitHub sous licence MIT.
Qui maintient modelright ?
Il est publié par homezloco. AI.tn le référence sans le maintenir ni le recommander.
Issu du registre MCP officiel. Être listé n'est pas une recommandation : vérifiez le code et les permissions d'un serveur avant de lui donner accès à vos données.