If your question isn't here, the companion knows the rest. Tap the orb in the corner — or open a ticket.
It takes a piece of content — a contract clause, a code diff, a claim, a memo — and runs it through a panel of frontier AI models. If the models disagree, BRIDGE orchestrates an adversarial debate until consensus or stable disagreement. You get back a signed verdict, applied corrections, and a confidence score.
Calling five models gives you five outputs. BRIDGE gives you one verified output — with adversarial debate, applied corrections, audit trail, signed consensus hash, and confidence-as-survival-score. The infrastructure isn't the hard part; the consensus + debate protocol is.
One complete protocol run: input dispatched to the model panel, adversarial debate if required, consensus reached, audit hash signed. Re-running on identical input within 24 hours is deduped from cache and doesn't count against your quota.
Verify takes a content artifact and returns corrections + verdict. Ask is conversational — you ask a question, the panel produces a verified answer. Same protocol under the hood; Ask just wraps it in a chat experience.
Yes. 100 verifications/month + 500 Ask queries on Tier 1, forever. No credit card. The free quota resets monthly. We don't sell your data or use it to train models — the free tier is sustained by the volume of paid customers, not by your data.
Yes. Verified .edu institutions and registered 501(c)(3) nonprofits get 50% off Builder and Team. Contact us with proof of status.
You're not blocked. We bill overage at $0.04/verification on Tier 2, $0.18 on Tier 3, and $0.02/$0.10 on Ask T2/T3 respectively. You can set a hard cap in Settings → Billing if you want to stop at quota instead.
20% off when billed annually on Builder and Team. Cancel any time during your annual term; we refund the unused months pro-rata.
No. By contract, neither BRIDGE nor any model panel member trains on Customer Content. This is enforced through zero-retention agreements with each provider. See Privacy for detail.
Default: US (us-east-1). On Team and Enterprise you can choose US, EU, or specific regions. On Enterprise you can configure full PII/PHI redaction before dispatch to the panel.
Customer Content is deleted within 30 days. The audit trail (hashes + metadata, no content) is retained per your plan's policy. Billing records are retained 7 years per tax law.
SOC 2 Type II in flight (Q3 2026). GDPR and CCPA compliant. HIPAA BAA available on Dedicated and Enterprise. SOX audit-trail aligned. ISO 27001 on roadmap. See Enterprise for the full compliance posture.
Under three minutes from API key to first call. POST /v1/verify with a bearer token and JSON body. SDKs in Python, JavaScript, Go, Ruby; betas in Rust and Java. See Developers for the quickstart.
Starter 60 req/min, Builder 600/min, Team 6,000/min, Enterprise custom. Bursts up to 2× sustained rate are absorbed.
Yes to all three. Plus MCP server (Claude Desktop, Cursor, Continue), Zapier app, and OpenAI / Copilot / Gemini plugins.
Team and Enterprise can replace any panel slot with a private model endpoint (OpenAI-compatible). The protocol — debate, consensus, audit — is identical regardless of which models you bring.
Yes, on Enterprise. Air-gap compatible. Local-only audit chain. Bring-your-own model panel. Federated update channel for protocol upgrades without external network access.
Yes, on Dedicated and Enterprise. Required for healthcare workloads handling PHI.
Cloud 99.9%, Dedicated 99.99%, with p95 latency under 5s. Penalty for missed SLA: 10% credit on Cloud, 30% on Dedicated, custom on Enterprise.
Median 28 days from first call to contract signature. Faster if you're cloud-only and standard SLA; longer for on-prem, air-gap, or regulated industries requiring security review.