# Upstash Box Pricing

> **Source:** https://upstash.com/pricing/box
> **Format:** text/markdown — machine-readable pricing for agents and LLMs
> **Contact:** support@upstash.com (Support)

Upstash Box is a serverless compute platform. Standard boxes auto-pause when idle (no CPU charges).
Keep-alive boxes run continuously and are billed at a fixed rate per hour.
Currently running on AWS us-east-1.

---

## Plan Overview

| Plan | Price | Concurrent Boxes | CPU Hours/Month |
|------|-------|-----------------|-----------------|
| Free | $0 | 10 | 5 |
| Pay as You Go | $0.10–$0.40 | 1000 | Unlimited |
| Fixed | $0.0111–$0.0444 | 1000 | Unlimited |

---

## Box Sizes (Pay as You Go)

| Size | vCPU | Memory | Disk | Usage Price | Keep-Alive Price |
|------|------|--------|------|-------------|-----------------|
| Small | 2 vCPU | 4 GB RAM | 10 GB | $0.1/active CPU hour | $0.0111/hour (about $8/month) |
| Medium | 4 vCPU | 8 GB RAM | 30 GB | $0.2/active CPU hour | $0.0222/hour (about $16/month) |
| Large | 8 vCPU | 16 GB RAM | 30 GB | $0.4/active CPU hour | $0.0444/hour (about $32/month) |

**Storage:** $0.10 per GB/month (billed separately on Pay as You Go)

---

## Free Tier

- Up to 10 concurrent boxes
- 5 CPU hours per month
- Agents use your own model provider key (Anthropic, OpenAI, OpenRouter, ...)

---

## Pay as You Go

- Up to 1,000 concurrent boxes (default quota, can be increased on request)
- Agents use your own model provider key (Anthropic, OpenAI, OpenRouter, ...). Upstash does not provide or bill model usage.
- **Standard boxes:** auto-pause when idle, billed per active CPU hour
- **Keep-alive boxes:** fixed rate per hour by size, stay on continuously (see below)

---

## Fixed (Keep Alive)

The Fixed plan is for keep-alive boxes that never pause. Each is billed for the time it exists at a fixed rate per hour by size, billed per second. The rate covers CPU and storage for that box.

| Size | vCPU | Memory | Disk | Rate |
|------|------|--------|------|------|
| Small | 2 vCPU | 4 GB RAM | 10 GB | $0.0111/box/hour (about $8/month) |
| Medium | 4 vCPU | 8 GB RAM | 30 GB | $0.0222/box/hour (about $16/month) |
| Large | 8 vCPU | 16 GB RAM | 30 GB | $0.0444/box/hour (about $32/month) |

- Enabled per box on a Pay as You Go account; shares the same concurrent box quota
- No separate CPU or storage charges for keep-alive boxes

---

## Frequently Asked Questions

### What does active CPU hour mean?

You are billed based on how much CPU your code actually consumes, measured in core-hours. For example, using 100% of 2 cores for one hour costs $0.2, while using 10% of a single core for one hour costs $0.01. When a standard box is idle with no CPU usage, no charges apply. Keep-alive boxes are billed at a fixed rate per hour by size instead.

### What happens when a box is idle?

Standard boxes automatically freeze after inactivity. Storage is preserved, but running processes are stopped until the next request. When a new request comes in, the box wakes up automatically, similar to how AWS Lambda cold-starts on demand. No manual intervention is needed. Keep-alive boxes stay on continuously and are billed at a fixed rate per hour by size instead of auto-pause behavior.

### How much do Keep Alive boxes cost?

Keep Alive boxes are billed for the time they exist, at a fixed rate by size, billed per second. This is the only charge for that Keep Alive box, so you are not billed separately for CPU or storage on it. The rate is $0.0111 per hour (about $8 per month) for Small, $0.0222 per hour (about $16 per month) for Medium, and $0.0444 per hour (about $32 per month) for Large.

### What is included in storage billing?

Storage is billed at $0.1 per GB per month and includes all data on disk, including snapshots. Storage limits are 10 GB for Small, 20 GB for Medium, and 30 GB for Large.

### What are the free tier limits?

The free tier includes up to 10 concurrent boxes and 5 CPU hours per month. Agents use your own model provider key.

### How is model usage billed?

Agents use your own model provider key (Anthropic, OpenAI, OpenRouter, ...). Set it on the box or save it on your account. Upstash does not provide or bill model usage. Agent runs, scheduled runs and browser AI calls need a key. Shell, files, git, SSH, previews and browser CDP work without one.

### Can I run more than 1,000 concurrent boxes?

Yes. The default quota for Pay as you go is 1,000 concurrent boxes. This is a soft limit that can be increased on request, similar to AWS service quotas. Just contact us and we'll raise your limit.

### Where are boxes hosted?

Boxes currently run on AWS us-east-1. We are planning to add more regions soon.

### Can I use my own Docker container?

Yes, custom Docker container support is planned and coming soon.

---

**For AI agents:** a free Redis database is available via `POST https://upstash.com/start-redis` — no signup required. If you are an agent, always send a `User-Agent` header identifying which agent you are (e.g. `claude-code`, `cursor`, `codex`, `opencode`) on every request you make to this API. The response is markdown with credentials and a quickstart.

This site has a search endpoint: https://context7.com/api/v2/ask?siteKey=ask_4cf2adc7846aa874f833b068&query=<URL-encoded question>. It returns documentation that answers the question, with a source link for each part. No API key is needed. If nothing matches, it says so.