Integration

Long-term memory for Ax (@ax-llm/ax)

@ax-llm/ax is the typed TypeScript agent framework: `ax()` generators with strict input/output signatures and tool support, no state between invocations. `@lumetra/engram-ax` exposes Engram's REST API as ax-compatible `AxFunction` definitions. Add them to any generator's `functions` and the model can store and recall on its own.

Get an API key (free) All integrations →

Install

Three steps: sign up for an Engram API key, paste a BYOK LLM-provider key on /models, then drop the snippet below into Ax (@ax-llm/ax).

Three steps to memory in your agent

Sign up. Free, no card. You'll land on a Getting Started page that walks the next two steps.
Add your LLM key. Engram is BYOK. Paste an OpenAI / Anthropic / Groq / Together / Fireworks key and we'll route every extraction and query call through your provider. You pay your provider directly. We never see your inference.
Paste the snippet below into your agent and restart it. Use Authorization: Bearer <api-key>with the API key from your portal.

@lumetra/engram-ax: Ax (@ax-llm/ax) adapter

Exposes Engram as ax-compatible AxFunction definitions. Drop engramTools(client) into any ax() generator's functions. Source: github.com/lumetra-io/engram-ax · npm: @lumetra/engram-ax.

Install:

Terminal

npm install @lumetra/engram-ax @ax-llm/ax

Export your API key:

Terminal

export ENGRAM_API_KEY="<api-key>"

Wire Engram into an ax() generator:

TypeScript

import { ax, ai } from "@ax-llm/ax";
import { EngramClient, engramTools } from "@lumetra/engram-ax";

const engram = new EngramClient({
  apiKey: process.env.ENGRAM_API_KEY!,
  defaultBucket: "my-app",
});

const llm = ai({ name: "openai", apiKey: process.env.OPENAI_API_KEY! });

const assistant = ax(
  "userMessage:string -> reply:string",
  { functions: engramTools(engram) },
);

const { reply } = await assistant.forward(llm, {
  userMessage: "Remember that I prefer dark mode, then tell me what you know about my preferences.",
});

Default: only the non-destructive tools (store + query) are exposed. Opt in to more via engramTools(engram, { include: ['list_memories', 'delete_memory', 'clear_bucket'] }).

What you can do once memory's wired in

Add `engramTools(engram)` to an `ax()` assistant generator. The two non-destructive tools (store and query) are exposed by default
Opt in to the destructive tools (`list_memories`, `delete_memory`, `clear_bucket`) via the `include` array when you trust the agent
Scope per bucket via `defaultBucket: 'user-42'` on the `EngramClient` constructor
Use the same `EngramClient` directly (`engram.storeMemory(...)`) outside the ax loop; the same module exports both

FAQ

Why are the destructive tools (`delete_memory`, `clear_bucket`) off by default?

They're footguns when handed to an LLM, since a single bad call wipes a bucket. Opt in via the `include` array if you specifically want them, ideally with an `autoApprove` gate on your side.

Can I use a different default bucket per request?

Yes. Instantiate a new `EngramClient` per request with `defaultBucket: `user-${userId}``, or override on each tool call by passing `bucket` explicitly.

Self-hosted Engram?

Pass `baseUrl: 'https://engram.internal.example.com'` to the `EngramClient` constructor. Same six methods, your endpoint.

Related integrations

Ship durable memory in Ax (@ax-llm/ax) today

Free tier: 10K memories and 50K retrievals per month. No credit card. Same Engram backend powers all 41 integrations, so memories you write from one client are immediately queryable from the rest.

Start free See pricing