---
title: "The Map — Context Infrastructure"
description: "One engine reached through several doors."
canonical: "https://press.getalchemystai.com/books/context-infrastructure-build-ai-that-understands-your-business/the-map"
last_updated: "2018-10-20"
book: "Context Infrastructure"
label: "Chapter 4"
---

# The Map

Alchemyst is one engine reached through several doors. Whichever door a team uses, every request lands on the same context and the same arithmetic, so a voice agent, a coding assistant and a support bot can share one source of truth.

**Every interface reaches the same engine and the same context.**

| Layer | Components |
| --- | --- |
| **Applications** — what users meet | Your agents (support, ops, copilots) · Kathan Voice OS (multilingual voice) · Example projects (GitAlchemy, CLI agent) |
| **Interfaces** — how you connect | REST API (OpenAPI-documented) · SDKs (TypeScript, Python) · MCP server (OAuth, memory tools) · Context Router (OpenAI-compatible) · Frameworks (Vercel AI, LangChain) · Platform UI (Playground, graph) |
| **Core APIs** — what you call | Context API (add, search, view, delete, async jobs) · Memory API (session memory: add, update, delete) · Steering API (search plus a grounded LLM answer) · Context Traces (sources, scores, rules per request) |
| **Engine** — what does the work | Context arithmetic (scope, exclude, rank) · Knowledge graph (references, inferences) · Managed infra (6 regions, zone choice) |

*Figure 4.1 · The Alchemyst product stack*

Read it bottom up: the engine stores and selects context, the core APIs expose it, the interfaces make it easy to call from your stack, and applications are what end users meet.

| Component | What it does | Who reaches for it |
| --- | --- | --- |
| **Context API** | Ingests documents (sync or queued), searches, views and deletes context | Every builder |
| **Memory API** | Stores and updates conversation history per `sessionId` | Builders of chat and voice agents |
| **Steering API** | Runs a search and returns an LLM answer grounded in what it found, shaped by a `steeringPrompt` | Teams that want answers, not chunks |
| **Context Traces** | Records which sources, scores and rules produced each context window | Builders debugging, reviewers auditing |
| **MCP server** | Exposes context and memory tools to Claude Desktop, Cursor, VS Code and any MCP client, signed in with OAuth 2.1 | Developers in their IDE |
| **Context Router** | An OpenAI-compatible proxy that injects scoped context into an existing agent's requests with no code changes | Teams with agents already in production |
| **Platform UI** | Playground, visual context graph, magic keys for read-only sharing, trace viewer | Builders and team leads |
| **Kathan Voice OS** | Multilingual voice agents running on the context layer, with write-back to systems of record | Operations teams running calling at scale |

---

- Previous: [Chapter 3: Six Ideas](https://press.getalchemystai.com/books/context-infrastructure-build-ai-that-understands-your-business/six-ideas.md)
- Contents: [Context Infrastructure](https://press.getalchemystai.com/books/context-infrastructure-build-ai-that-understands-your-business.md)
- Next: [Chapter 5: Ten Minutes](https://press.getalchemystai.com/books/context-infrastructure-build-ai-that-understands-your-business/ten-minutes.md)
