Multi-model is the default.
Product teams route across providers for quality, cost, and latency. Single-provider assumptions no longer hold.
MVP AI infrastructure LLM gateway
QK Gatewai
Multi-provider inference without the integration sprawl: deterministic tests, policy-aware routing, and one request trail.
- Deterministic fixtures
- Policy-aware relay
- 18 catalogued adapters
- Usage and audit
MVP Pre-revenue · founder-led · internal dogfooding only
18provider adapters catalogued
4native API formats
30keyless dev/test fixtures
1internal product live on Gatewai
THE OPPORTUNITY
Every AI product is becoming a multi-provider product.
Provider integrations differ in request formats, credentials, failure behavior, and operational details. Teams must absorb those differences while deciding how retries, fallback, usage, quotas, and access rules should work.
Testing against live services requires provider keys, depends on external availability, and incurs usage charges. Without deterministic fixtures and a clear request trail, failures are hard to reproduce and spend is hard to explain.
Inference is a budget line.
Usage, quotas, and chargeback need enforcement before the request, not a dashboard after it.
Provider churn is operational.
Models, formats, limits, and terms change. The integration layer has to absorb the change.
Initial buyer hypothesis. Platform and API teams at AI product companies and studios running multiple providers. The economic buyer is the engineering leader accountable for inference spend and reliability.
THE PRODUCT
One relay for access, policy, and proof.
Gatewai combines deterministic keyless fixtures for development with a live relay that checks provider mappings, applies routing and policy, records usage, and leaves an audit trail.
Keep provider-specific handling out of each product
The descriptor catalogue represents 18 provider adapters and accepts OpenAI Chat, OpenAI Responses, Anthropic Claude, and Google Gemini request formats. This is catalog scope, not a claim that every provider has been live-qualified; broader compatibility validation remains in progress.
Policy inside the relay
A 10-stage pipeline supports policy-aware channel selection, guards, fail-closed provider checks, usage and quota controls, and request tracing.
Workspaces and accountability
Workspace access controls, audit history, and operator surfaces make provider activity and governance easier to inspect.
Reproduce request paths without live credentials
Thirty deterministic, keyless fixtures support repeatable development and tests. The offline path covers a stub-provider request, an OpenAI-shaped response, usage recording, and the settled request-log entry.
Developer entry point: the QK Gatewai Community Edition packages the zero-credential local path for individual developers.
PRODUCT SURFACE
A working operator console—not just a routing concept.
Explore the dashboard, connectivity, offline samples, and governance surfaces. Open any image to browse the full set of 25 captured console screens in the lightbox.
EVIDENCE & ROADMAP
What is built—and what is still planned.
IMPLEMENTEDBuilt
- Provider and media relay. 18 catalogued adapters and four wire formats; live-provider compatibility is still being qualified. Image, video, and audio job routes are part of the relay.
- Routing and resilience. A 10-stage pipeline with channel groups, priority failover, model fallback, health monitoring, and per-tenant rate controls. Missing mappings fail closed.
- Governance and access. Guards, workspaces, role-based access, scoped tokens, quotas, audit history, and encrypted credentials around the relay.
- Operations. A console for channels, models, pricing, request and usage logs, audit, health, task monitoring, and billing operations.
- Repeatable evaluation. Thirty keyless fixtures and a zero-credential offline path through usage and the settled request log.
- Internal integration. Novelino routes model and media calls through Gatewai; portfolio use, not external adoption.
NEXTPlanned
- Commercial usage metering. Per-model token and media markup, with quota holds before relay and settlement afterward.
- Self-serve storefront. Public signup, API-key provisioning, usage dashboards, and pay-as-you-go credit wallets.
- Plans and payments. Recurring tiers, prepaid credits and top-ups, invoicing, and refunds.
- Team controls. Seats, pooled organization budgets, per-workspace spend caps, and role-gated model access.
- Enterprise and scale. Dedicated single-tenant gateways, BYOK governance, custom SLAs, volume commitments, and multi-region failover.
- Broader qualification. Closing live-provider compatibility gaps across relay, media, account, billing, and operations.
Billing and plan-management components exist internally, but resale is off by default and some subscription flows remain incomplete. The commercial offer is not generally available.
Internal dogfooding, not external traction. Novelino routes its model and media calls through Gatewai. No external customer adoption, revenue, or measured impact is claimed.
COMMERCIAL THESIS
Start where integration pain is immediate. Expand into governance.
The wedge is developer experience: compatible APIs and deterministic fixtures. The expansion is operational control: routing policy, quotas, audit, and workspaces. Monetization follows usage and assurance.
Compatible APIs lower the integration barrier.
Developers connect through familiar request formats and can test without a provider key; the gateway absorbs provider-specific differences.
Move from connectivity to operational control.
Routing policy, guards, workspaces, quotas, traces, and audit build on the same relay as teams add providers and use cases.
Monetize usage and higher-assurance needs.
Planned directions: usage-based text and media markup, team subscriptions, prepaid credits, self-serve access, and dedicated enterprise gateways.
Where Gatewai fits
| Category | Typical strength | What it leaves open |
|---|---|---|
| Single-provider SDKs | Fastest path to one provider | Per-app integration; no central governance or request trail |
| Routing proxies | Multi-provider routing and fallback | Test determinism, policy depth, and audit are separate concerns |
| Observability tools | Traces, cost visibility, and evaluation | Visibility without enforcement or routing |
| Cloud and edge gateways | Scale, security, and network integration | Provider-specific semantics and keyless development fixtures |
| Gatewai | Deterministic fixtures plus policy, usage, and audit in one self-hostable relay | Live-provider qualification and commercial packaging are in progress |
What must be true
- Willingness to pay. Teams already pay for inference; the open question is whether they pay a markup for control.
- Gross margin. Relay markup must cover infrastructure, support, and provider costs.
- Self-serve conversion. The developer entry point must convert to paid usage without high-touch sales.
- Design-partner pull. Three external teams must adopt before GA and before resale is enabled.
THE ASK
A 90-day plan to turn a working MVP into a commercial signal.
Seeking a milestone-based pre-seed partnership. Raise amount, runway, and allocation are in the funder brief.
01Days 1–30
Qualify five providers end-to-end.
Close live-provider gaps, publish a repeatable provider benchmark, and finish a security review.
02Days 31–60
Onboard three design partners.
Convert Novelino into a measured case study and test pricing with external teams.
03Days 61–90
Turn on metered billing and self-serve.
Enable usage settlement, signup, and API-key provisioning; publish the first external usage report.
Use of capital. Engineering, design-partner onboarding, security and compliance, and go-to-market. Raise amount, runway, and detailed allocation are in the funder brief.
TEAM & RISKS
Founder-led execution, with the risks named.
FOUNDER-LED EXECUTION
Product, architecture, and delivery in one seat.
Product direction, architecture, and delivery stay with the founder and technical lead. QuadKit was built without external funding; funders work directly with the person accountable for execution.
QuadKit supplies reusable foundations; coding agents extend implementation capacity; the founder reviews architecture, tests, and acceptance. Capital goes to product milestones, not rebuilding the platform.
Risks and mitigations
- Provider terms and resale. Start with BYOK and self-hosted deployments; enable resale only where provider terms allow.
- Commoditization by providers and clouds. Compete on test determinism and governance workflow, not raw routing.
- Security and key custody. Encrypted credentials, scoped tokens, audit, and least privilege; fund a security review in the first 30 days.
- Open-source alternatives. A public Community Edition brings the developer entry point; paid governance and enterprise deployments are the commercial layer.
- Solo-founder execution. Documented contracts, tests, and architecture; the first engineering hire follows design-partner validation.