Can't find what you need? support@blastoff.live
Blastoff is an agent tooling platform. One of first tool lets you point your AI agent or LLM endpoint, upload a dataset of inputs and expected outputs, and Blastoff fires every row, scores the responses, and shows you pass/fail results in real time.
JSON — an array of objects, each with an `input` field and an `expected_output` field. You can also add `assertion` (array of strings) and `threshold` (0.0–1.0) per row. Download the sample dataset from your suite page for an example.
Blastoff's built-in LLM judge scores your endpoint's responses against expected outputs without needing your own API key. Free users get 1 judge run/month; Pro users get 15.
You can add your own Anthropic or OpenAI API key in Settings. Runs using your key are charged to your account directly and don't count against Blastoff judge limits. Free users get 3 BYO runs/month; Pro users get unlimited.
The security scan fires 165 adversarial prompts across 9 attack categories (prompt injection, jailbreaks, encoding attacks, persona manipulation, and more) at your endpoint and reports which ones were EXPLOITED vs DEFLECTED. Available on Pro.
Connect your GitHub repo in the setup wizard. Blastoff creates a GitHub Check Run on every pull request and blocks merges if your eval score drops below your configured threshold.
Go to Settings → Manage Billing. You'll be taken to the Stripe billing portal where you can cancel. You keep Pro access until the end of the billing period.
Yes — API keys are encrypted with Fernet symmetric encryption before being stored and are never logged or exposed in responses.
Still need help?
Email us directly — we read and respond to every message.
support@blastoff.live