# TokenZero (tokenzero.sh)

Routes LLM API calls through a proxy that adds prompt caching, exact-response caching and a signed savings ledger.

- Atlas card: https://inite.ai/en/atlas/tokenzero.sh
- Site: https://tokenzero.sh
- For: Developers and teams paying OpenAI, Anthropic or Gemini API bills who want to cut token spend.
- Categories: llmops
- Use not measured/100 · alive 100/100

## Used for

- route LLM API calls through a gateway to reduce spend
- make prompts cache-friendly without changing their content
- scan a codebase for uncached prompts and costly model choices

## Price

- from 1790 USD/year, free plan (read 2026-10-29 on https://tokenzero.sh/#pricing)

## For developers

- OpenAPI: https://tokenzero.sh/openapi.json (30 paths)
- API docs: https://tokenzero.sh/openapi.json
- MCP server: io.github.tokenzero-ai/tokenzero (github; open on 2026-10-29, 5 tools)
- MCP server: sh.tokenzero/mcp (namespace; open on 2026-10-29, 5 tools)
- Read 2026-10-29

## MCP server, measured daily

Server: https://api.tokenzero.sh/mcp, checked every day since 2026-10-29 (4 checks).

- Answered: measuring, 4 daily checks so far
- Median first answer: 224 ms
- Sign-in: none
- Context its tool definitions take: about 1066 tokens (6 tools)
- Tool definitions: A, 92/100
- Failing "params_described": 3 of 6 tools
- Method: https://inite.ai/en/atlas/mcp/report

## Your data

- Certifications stated: none
- Read 2026-10-29

---

Source: INITE Atlas, https://inite.ai/en/atlas/tokenzero.sh. JSON: https://inite.ai/api/atlas/tokenzero.sh. MCP: https://inite.ai/api/mcp (atlas_lookup).