Security & Privacy

🔒 Encryption

• TLS 1.3 on all connections via Cloudflare — no plaintext traffic accepted
• HSTS enabled with preload — browsers always use HTTPS
• API keys and credentials encrypted at rest — never stored in plaintext
• Stripe handles all card data — we never see or store card numbers

👁 Data Handling — What We Store

We follow a zero-payload-retention policy:

• Request bodies — NOT stored. Your prompts, text, and data are processed in memory and discarded.
• Response bodies — NOT stored. AI outputs are returned to you and not retained.
• Billing metadata only — we log: timestamp, endpoint called, token count (for metered billing), and API key ID. No content.
• Agent memory — stored only if you explicitly use /memory endpoints. You control it and can delete it anytime.
• IP addresses — used only for rate limiting, not sold or shared.

🛡 Code Sandbox

When you use the /code/run endpoint, your code runs in a restricted sandbox:

• AST-validated before execution — dangerous patterns rejected at parse time
• No filesystem access — os, sys, pathlib, shutil blocked
• No network access — socket, requests, urllib, httpx blocked
• No process spawning — subprocess, multiprocessing, threading blocked
• Blocked builtins: eval, exec, compile, open, __import__, getattr, setattr
• Execution timeout enforced — runaway code is killed

🛡 SSRF Protection

All outbound HTTP requests (scraping, webhooks, API catalog) pass through SSRF validation:

• Private IP ranges blocked (10.x, 172.16-31.x, 192.168.x, 127.x, ::1)
• Link-local and metadata endpoints blocked (169.254.x, cloud metadata)
• DNS rebinding protection — resolved IPs checked against block list

💰 Payment Security

• Stripe handles all card payments — PCI DSS compliant, we never touch card data
• x402 USDC payments verified on-chain via Coinbase facilitator on Base Mainnet
• Automatic refund credits — if a paid request returns a 5xx error, you get a refund credit automatically (returned in X-Refund-Credit header)
• API keys are revocable — contact us to deactivate a compromised key
• Request correlation via X-Request-ID header on every response

🤖 For AI Agents

Machine-readable security signals are embedded in every interaction:

• /.well-known/agent.json includes a security object with data handling policies
• Every 402 response includes security field confirming data retention policy
• /.well-known/security.txt follows RFC 9116
• X-Payment-Receipt header confirms payment was processed
• X-Refund-Credit header on 5xx after payment — automatic compensation

✉ Reporting Vulnerabilities

Found a security issue? Email [email protected] with details. We take all reports seriously and will respond within 24 hours.