Provider connections
Connect BYOK credentials, managed Platform Access, health checks, and provider priority.
Route AI requests across providers, fall back when one breaks, cap spend per key, and trace every call in real time.
LIVE ROUTING FLOW
A request enters one API, LatentKit evaluates policy, credits, provider health, key caps, and fallback state, then routes to the best eligible provider.
CONTROL SURFACE
LatentKit turns provider sprawl into product controls your team can operate.
Connect BYOK credentials, managed Platform Access, health checks, and provider priority.
Control strategy, fallback depth, response profile, route order, and publish state.
Issue reveal-once keys with monthly caps, expirations, rotation, and instant revoke.
Separate BYOK provider billing from managed-credit usage, margin rules, and reloads.
Inspect candidates, selected route, fallback chain, cost, latency, cache, and request ID.
Test chat, images, and embeddings through the same policy before production traffic.
Apply prompt controls, trace privacy settings, PII handling, and anomaly visibility.
Manage roles, audit events, branded domains, white-label surfaces, and support workflows.
SETUP FLOW
01
LatentKit prepares the default app, route, key, and quickstart path when Platform Access is enabled.
02
Bring your own provider keys or use managed Platform Access where it is available for the route.
03
Set endpoint capability, response profile, fallback limit, provider order, and rollout timing.
04
Create a reveal-once app key with caps and expiration, then test the live policy in Playground.
05
Keep the OpenAI request shape and point production traffic at the LatentKit gateway.
STRATEGY LAB
Same policy, different strategy. See how priority, cost, availability, and balanced mode distribute traffic.
Priority mode keeps production simple: one primary route, with the rest ready as fallback.
POLICY PUBLISHING
Tune response profile, eligibility, fallback depth, and canary rollout from the console while your app keeps one endpoint.
Preview
Publish v17 with priority mode, balanced response profile, OpenAI as managed fallback, and Gemini excluded while degraded.
LIVE OBSERVABILITY
Logs, route candidates, fallback attempts, cost, latency, cache, guardrails, and request IDs.
Request trace
Selected first healthy BYOK route. No guardrail match. Stored with normal trace metadata.
DEVELOPER EXPERIENCE
Use the official Python and JavaScript SDKs, cURL, fetch, or anything that speaks OpenAI-compatible requests.
from latentkit import LatentKit
with LatentKit(api_key="USE_YOUR_LATENTKIT_KEY_HERE") as client:
response = client.chat.create(
messages=[{"role": "user", "content": "Route this"}],
response_profile="balanced",
)
print(response["content"])
ENTERPRISE OPERATIONS
Custom domains, white-label dashboard, roles, audit logs, guardrails, and support options.
Policy v12 scheduled canary at 12%
Key lk_live_91af rotated by Tenant Admin
Custom domain certificate healthy
PRICING
Free
Unlimited BYOK under fair use, no LatentKit usage markup, one app, one policy, and a managed trial path.
Starter
5 apps, 8 provider connections, 10 policies, 5 members, a $5 activation credit, and Platform Access at provider cost + 5%.
Company
Custom client workspaces, domains, branding, audit/export, governance, limits, and support.
WHY LATENTKIT
Flexible, but every fallback, budget, trace, and provider change becomes your infrastructure.
Useful for one provider family, weaker when your app needs policy control across many providers.
One API layer for routing, provider connections, budgets, traces, credits, and team operations.
Connect a provider, publish a policy, ship in under 15 minutes.