The institutional memory your AI agents need to operate.
Alchemyst AI is the context backbone that keeps every agent's knowledge current, traceable and semantically consistent across your entire organisation through a single API.
The model is replaceable.
Your institutional context isn't.
Models are commoditizing fast. Durable advantage comes from a context layer that operationalizes your business intelligence and stays yours no matter which model you run it on.
Models will keep changing. Your institutional context is the asset that compounds - so it should belong to you, not to whichever model you happen to run today.
Read the Context Thesis →
Switch models freely. Keep your context sovereign.
Every model swap normally resets your agent's memory. Alchemyst decouples what your organization knows from whichever model reasons over it, so institutional context stays continuous across every upgrade or multi-model setup.
Operationalize intelligence that runs your day-to-day.
This isn't a smarter chatbot. It's a context layer that turns what your organization knows into agents that run sales, support, ops, and research at scale — every decision traceable, every agent on the same source of truth.
One sovereign context layer, any model, agents that operate
A context layer that keeps your AI current, traceable, and semantically consistent.
One API call. Context arithmetic over your institutional knowledge graph. Every decision traceable back to its source - without managing a single vector database or graph store.
Context Arithmetic - the core primitive
Context arithmetic is the foundational primitive: dynamic set algebra over meaning, computed at query time. Instead of naïve top-K similarity, Alchemyst intersects to narrow scope, unions to widen recall, subtracts superseded or out-of-scope content, and ranks what remains - so only the right context survives into the window.
// Set algebra over meaning, at query time
const window = alchemyst.context.search({
query: userMessage,
groupName: ["sales", "emea"], // ∩ narrow scope
metadata: { version: "v2" }, // ∩ filter
});
// − superseded / deduped → rank → top-KInstitutional knowledge graph + context traces
What you store is an institutional knowledge graph of your organization's context, fully traceable. Memory isn't three hard-coded layers - by applying context arithmetic over the graph you can derive the behaviors people expect from memory: recall what happened, resolve what it means, and inform how to act. The memory types are outcomes of the primitive, not separate modules.
// One graph + arithmetic → derived "memories"
const whatHappened = ctx.search({ groupName: [session_id] });
const whatItMeans = ctx.search({ query: term })
.subtract(deprecated);
// "how to act" falls out of ranked, in-scope contextContext Traces for full auditability
Every agent decision is traceable back to the exact context it had - at a query level. Not a summary, but the exact data points, scores, and rules that went into the model's context window. Debug in minutes, not days.
const trace = await alchemyst.trace.get(
session_id, turn_id
);
// Returns: sources[], scores[], rules_applied[]
// Pairs with Euphony for visual debuggingSemantic consensus enforcement
Define canonical term definitions at the org level. When "revenue" means different things to different teams, Alchemyst resolves the ambiguity before it reaches the model.
await alchemyst.ontology.define({
term: "revenue",
canonical: "ARR as reported to board",
aliases: ["sales", "bookings", "ARR"],
owner: "finance",
updated_at: new Date()
});Illustrative · accuracy vs conversation context
Example Use Case
How do you debug what an agent can't see? Context Tracing with OpenAI Euphony
Pairing Alchemyst's Context Traces with Euphony - OpenAI's open-source conversation visualizer - creates an end-to-end debugging workflow. Every agent failure is now diagnosable in minutes: was it a retrieval problem, a configuration problem, or a model problem?
Give your AI agents the memory they deserve.
Join developers building the next generation of AI products with persistent, auditable context. Free tier available — no credit card required.