Skip to main content
All posts
ProductSandboxBYOKSecurityLaunch

Try Before You Sign Up — Our New Free Sandbox Lets You Test With Your Own API Key

6 min readby Clawship Team

The number-one reason people bounce from a SaaS landing page: they can’t try it without signing up. Today we’re fixing that.

clawship.app/try is a free, zero-signup sandbox. Paste your own API key, pick a personality template, and you’re chatting with a live AI assistant in under 30 seconds. No account. No credit card. No trial that expires in 14 days.

Here’s how it works, how we keep your key safe, and why we think every SaaS should let people test the product before asking for an email address.

What the Sandbox Does

The sandbox gives you a full Clawship experience — the same engine that powers deployed instances — but scoped to a single browser session. Three steps:

1

Paste your key

Supports Anthropic, OpenAI, Google Gemini, xAI (Grok), and DeepSeek. We validate the key with a single low-cost API call before starting.

2

Pick a personality

Choose from pre-built templates — developer helper, writing coach, data analyst, and more — so the conversation is immediately useful.

3

Start chatting

You're talking to a real AI assistant. Same engine, same quality. Up to 50 messages over 30 minutes.

When your session ends — or you close the tab — everything is cleaned up. The session is deleted from memory and your API key is gone. There is nothing to cancel, nothing to remember.

Why Bring Your Own Key?

We considered giving everyone free credits. It’s the obvious path. But it creates three problems:

  1. Abuse. Free credits attract bot farms and key scrapers. You end up spending more on fraud prevention than on the credits themselves.
  2. Stale impressions. Rate-limited free tiers feel broken. If the first experience is “please wait, you’ve exceeded your limit,” the user won’t come back.
  3. Wrong signal. Someone willing to paste their own API key is a real potential customer. They already have an AI workflow. They’re evaluating platforms. That’s exactly who we want to impress.

BYOK self-selects for serious users while keeping our costs at zero. The user pays their own provider directly — we just provide the infrastructure to route their key through a real assistant.

How We Keep Your Key Safe

This is the part most people care about — and should care about. Here’s the full picture:

Keys never touch a database

Your API key exists only in server-side memory for the duration of your session. When the session ends, the entry is deleted. There is no persistence layer.

Keys never reach the browser after submission

After you paste your key in the form, it's sent once over HTTPS to the backend. From that point on, the frontend only holds a session token — never the key itself.

Sessions auto-expire

Every session has a hard 30-minute TTL and a 50-message cap. When either limit is hit, the session is destroyed and the key is wiped from memory.

We chose REST proxying over direct WebSocket connections specifically so your key stays server-side. Every message you send goes through our backend, which injects the key into the AI provider request and strips it from the response. The browser never sees the key again after the initial form submission.

Under the Hood

The sandbox uses the same infrastructure that powers paid instances. Here’s the high-level flow:

sandbox session lifecycle
Browser → Clawship backend (API key + provider + personality)
  ↓
Validate key — single low-cost call to the provider to confirm the key is valid
  ↓
Create session — provision an agent with your chosen personality, store key in server memory only
  ↓
Chat — each message is proxied through our backend, which injects the key server-side
  ↓
Cleanup — session end (manual, timeout, or message cap) → agent removed, key wiped from memory

The key design decision was server-side proxying. A direct connection from the browser to the AI provider would require exposing the API key to the client. By proxying every message through our backend, the key stays server-side for the entire session.

Sandbox sessions use the same engine that powers paid instances, so you get production-quality responses — not a degraded demo mode.

Fair Use Limits

The sandbox runs on shared infrastructure, so we cap sessions to keep things fair:

  • 50 messages per session — enough to have a real conversation, not enough to run a production workload.
  • 30 minutes per session — after that, the session auto-terminates and the slot is released.
  • Capacity indicator — the UI shows how many sandbox slots are available in real-time. If we’re at capacity, you know immediately instead of hitting a cryptic error.

These limits are generous enough to evaluate the product seriously. Most test sessions last 5-10 minutes and 10-20 messages. If you need more, that’s a sign Clawship is working for you — sign up and deploy your own instance with no session limits.

Supported Providers

The sandbox supports five AI providers at launch:

ProviderModelsGet a Key
AnthropicOpus 4.6, Sonnet 4Get key →
OpenAIGPT-4.1, o3Get key →
GoogleGemini 2.5 Pro, FlashGet key →
xAIGrok 3Get key →
DeepSeekDeepSeek V3, R1Get key →

Don’t have an API key yet? Most providers offer free credits when you create an account. Google AI Studio gives $300 in free credits; Anthropic and OpenAI both offer initial usage allowances.

Try It Now

The sandbox is live. No waitlist, no beta invite, no “request access” form.

Try the Sandbox30 seconds to your first message. Your key, our platform.

If you run into capacity limits, check back in a few minutes — sessions are short and slots free up quickly. And if you like what you see, create an account to deploy a permanent assistant across WhatsApp, Telegram, Discord, Slack, and 10+ more channels.