Engineered for trust
- Transactions with row locks, atomic operations
- URL dedup and expiration policies
- Automatic retries and failover
- 3,000+ tests passing
About
We do one thing: trustworthy AI reports.
Stable ingestion, auditable pipelines, predictable billing — that is the whole obsession.
Principles
Three engineering commitments we hold the line on.
Trustworthy, operable, scalable — each backed by facts you can verify.
Operable
- Complete logs — every attempt recorded
- RETRYING and FAILED kept distinct
- Cookies and sessions auto-redacted
- Real-time instance monitoring
Scalable
- BullMQ async processing, horizontal scale-out
- Full multi-tenant data isolation
- Collectors, cleaners, and generators decoupled
- OpenAI SDK-compatible interface
Trust is not a slogan. Trust is a failed report that can explain exactly why it failed.
Transparency
Transparent architecture, auditable by design.
Every collection, every LLM call, and every charge has its own log and a traceable chain of evidence. Security audit scope and remediation results are published on the trust page; enterprise customers can request a detailed security brief.
- Task queues
- 13
- Tests
- 3,000+
- Key encryption
- AES-256
- Stack
- TypeScript