# Kourier (kourier.sh)

Serves open-weight language models through an OpenAI-compatible API endpoint running on the company's own GPU clusters.

- Atlas card: https://inite.ai/en/atlas/kourier.sh
- Site: https://kourier.sh
- For: Developers and small teams running agentic coding tools who want flat-rate inference without per-token charges
- Categories: llmops
- Use 0/100 · alive 100/100

## Used for

- Connect Cursor or Zed to an OpenAI-compatible endpoint
- Run long agentic coding sessions without per-token billing
- Serve open-weight DeepSeek v4.1 Flash models through an API

## Price

- from 20 USD/month, no free plan (read 2026-10-21 on https://kourier.sh/pricing)

## For developers

- OpenAPI: https://kourier.sh/openapi.json (5 paths)
- MCP server: sh.kourier/info (namespace; open on 2026-10-21, 8 tools)
- Status page: https://kourier.sh/status
- Read 2026-10-21

## MCP server, measured daily

Server: https://kourier.sh/mcp, checked every day since 2026-10-21 (2 checks).

- Answered: measuring, 2 daily checks so far
- Median first answer: 220 ms
- Sign-in: none
- Context its tool definitions take: about 993 tokens (8 tools)
- Tool definitions: A, 96/100
- Failing "params_described": 2 of 8 tools
- Warning "open_writes": create_checkout_link
- Method: https://inite.ai/en/atlas/mcp/report

## Your data

- Certifications stated: none
- Training on your data: no (https://kourier.sh/privacy-policy)
- Read 2026-10-21

---

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