# A7OM (ATOM)

> The Global Price Benchmark for AI Inference. Independent. Transparent.

ATOM is the world's first institutional-grade pricing intelligence platform for AI inference. We track 1,600+ SKUs across 40+ vendors and 6 modalities, publishing the AIPI (ATOM Inference Price Index) — a family of 25 chained matched-model price benchmarks covering global and regional markets.

## Products

- [ATOM MCP](https://a7om.com/mcp): AI inference pricing intelligence delivered as a native MCP tool for AI agents. 4 free tools, 4 additional tools with MCP PRO ($49/mo). Compatible with Claude, Cursor, Windsurf, and any MCP-compatible client.
- [ATOM Terminal](https://a7om.com/terminal): Institutional-grade dashboards for analysts and FinOps teams. 25 indexes, vendor benchmarking, competitiveness rankings, model-level data, complete time series, and daily alerts.
- [ATOM Feed](https://a7om.com/feed): Structured inference pricing data delivered direct to enterprise systems. 32 fields per SKU. For engineering and quant teams.

## Coverage

- 1,600+ SKUs indexed weekly
- 40+ vendors from hyperscalers to specialized inference providers
- 6 modalities: Text, Image, Audio, Video, Voice, Multimodal
- 25 AIPI benchmark indexes (global and regional)
- 32 fields per SKU including model intelligence (context window, tool calling, open-source flag, parameter count, model family)

## Methodology

ATOM uses a deterministic, zero-variance extraction methodology. Pricing data is collected directly from vendor pricing pages using vendor-specific parsers, normalized to standard units (per 1,000 tokens for text, per image, per minute, per second, per 1,000 characters), and indexed using a chained matched-model methodology that eliminates composition bias. All indexes exclude reseller/marketplace pricing (e.g., OpenRouter) to reflect primary market prices. Full methodology is published at https://a7om.com/methodology.

## MCP Server

ATOM's MCP server provides programmatic access to inference pricing data for AI agents:

- Server URL: https://atom-mcp-server-production.up.railway.app/mcp
- Transport: HTTP SSE and stdio
- Authentication: API key via x-api-key header (no key = free tier with 4 tools, valid key = full access with 8 tools)
- npm: https://www.npmjs.com/package/atom-mcp-server
- GitHub: https://github.com/A7OM-AI/atom-mcp-server

### MCP Tools

- list_vendors: Returns all 40+ tracked vendors with type and region
- get_kpis: Returns 6 market-level KPIs (output premium, caching savings, open-source advantage, context cost curve, caching availability, size spread)
- get_index_benchmarks: Returns AIPI chained matched-model price benchmarks across 25 indexes
- get_market_stats: Returns aggregate market statistics (SKU count, model count, vendor count, modality breakdown)
- search_models (PRO): Filter and rank models by modality, context window, tool calling, vendor, region, price range
- get_model_detail (PRO): Full specs and pricing for a single model across all vendors
- compare_prices (PRO): Cross-vendor price comparison with percentage differentials
- get_vendor_catalog (PRO): Complete catalog for a specific vendor

## Key Market KPIs (as of March 2026)

- Output Premium: 3.82x (output tokens cost ~4x input tokens)
- Caching Savings: 68.8% (average discount for cached input)
- Open Source Advantage: 82.5% (open-source models cheaper than proprietary)
- Context Cost Curve: 3.8x (large context models cost ~4x standard)
- Caching Availability: 22.6% (share of SKUs offering cached pricing)
- Size Spread: 5.0x (price ratio between large and small models)

## Company

ATOM is operated by Member One EOOD (Bulgaria). Founded by Stamatios Kanellakis, an information economist with 25+ years of experience in pricing strategy and monetization across enterprise technology markets.

## Links

- Website: https://a7om.com
- MCP Product Page: https://a7om.com/mcp
- Terminal Product Page: https://a7om.com/terminal
- Feed Product Page: https://a7om.com/feed
- Methodology: https://a7om.com/methodology
- LinkedIn: https://www.linkedin.com/company/a7om
- X/Twitter: https://x.com/ATOMInference
- Contact: https://a7om.com/contact