Global edge distribution
Serve cached responses from the nearest region to your user.
Agent-native edge caching
Caching for AI agents: faster, cheaper, visible.
Offload repetitive agent traffic at the edge to reduce latency and inference costs with a lightweight integration.
Request path
Illustrative cache behavior
Agent request
chat.completions.create()
Edge cache
Check for a reusable response
Model provider
Run only when no reusable response is found
Reuse repeated work while keeping cache behavior observable.
Multi-layer caching
Serve each request from the right layer.
From edge CDNs to regional Redis and local memory, Alchymos routes agent traffic to the fastest available layer.
Selected layer · CDN edge
Serve common, read-heavy responses close to users for low-latency global access.
A lightweight integration
Add caching to the clients you already use.
Connect an existing OpenAI client and keep working with familiar APIs. Start with the SDK, then configure the caching behavior that fits your workflow.
See SDK documentationTypeScript
import OpenAI from "openai";
import { Alchymos } from "@alchymos/openai";
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });
const alchymos = new Alchymos(process.env.ALCHYMOS_API_KEY);
alchymos.withLlm(openai);
const response = await openai.chat.completions.create(params);Built for control
Caching that fits your workflow.
Choose how requests are scoped, routed, inspected, and refreshed.
Authenticated caching
Scope cached data to organizations, licenses, or individual users.
Fine-grained rules
Configure behavior by tool, parameters, or request type.
Automatic invalidation
Purge entries when upstream data changes.
Purge API
Clear cache entries manually or programmatically.
Metrics and analytics
Review cache hits, latency, and tokens saved.
Cache debugging
Inspect headers to understand why requests are cached or bypassed.
Performance controls
Set performance targets for enterprise workloads.
Explore the SDK to see integration examples, or talk with us about caching for your agent workflows.