Integration

Long-term memory for Inngest AgentKit

Inngest AgentKit builds durable agents that survive restarts, retries, and queue shifts via Inngest's step engine, but the model's working memory still resets on every fresh run. `@lumetra/engram-agentkit` ships six AgentKit-compatible tools through `createEngramTools({...})`. Pass them to `createAgent` and the agent's durability-across-restarts now extends to what it actually knows, not just its execution graph.

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 Inngest AgentKit.

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-agentkit: six tools for any createAgent

createEngramTools({ ... }) returns AgentKit-compatible tools you pass into createAgent. Pairs Inngest's durable step engine with durable memory. Source: github.com/lumetra-io/engram-agentkit.

Install:

Terminal

npm install @lumetra/engram-agentkit @inngest/agent-kit

Wire the tools into an AgentKit agent:

TypeScript

import { createAgent, anthropic } from "@inngest/agent-kit";
import { createEngramTools } from "@lumetra/engram-agentkit";

const agent = createAgent({
  name: "researcher",
  system: "Use engram_query_memory before answering. Store stable facts.",
  model: anthropic({ model: "claude-sonnet-4-5" }),
  tools: createEngramTools({
    apiKey: process.env.ENGRAM_API_KEY!,
    bucket: "my-app",
  }),
  // AgentKit 0.13.2 Anthropic adapter bug — keep maxIter at 1 for tool flows.
  maxIter: 1,
});

For self-hosted Engram, pass baseUrl: "https://engram.internal.example.com" to createEngramTools.

What you can do once memory's wired in

`tools: createEngramTools({ apiKey, bucket: 'my-app' })` on a `createAgent({...})` and the agent gets store/query/list/delete out of the box
Pair Inngest's `step.run` retries with Engram's idempotent stores — a retried step can re-check memory instead of double-writing
Per-user buckets via `createEngramTools({ apiKey, bucket: `user-${userId}` })` constructed per event
Use the exported `EngramClient` directly inside a non-agent Inngest function for ETL-style memory ingestion

FAQ

Why is `maxIter: 1` recommended in the README?

AgentKit 0.13.2's Anthropic adapter throws `tool_use blocks can only be in assistant messages` when `maxIter >= 2` mid-loop. The bug is in how AgentKit serializes the tool-result message back to Claude — it's upstream, not in `@lumetra/engram-agentkit`. Keep `maxIter: 1` for Anthropic + tool flows until AgentKit fixes it; OpenAI / Gemini loops are unaffected.

Can I bring my own `EngramClient`?

Yes. `createEngramTools({ client })` accepts a pre-built `EngramClient` instead of `apiKey`. Handy when you want one shared client across Inngest functions, or a custom `fetch` for tracing.

Self-hosted Engram?

Pass `baseUrl: 'https://engram.internal.example.com'` to `createEngramTools(...)` (or to the underlying `EngramClient` constructor). Same six tools, your endpoint.

Related integrations

Ship durable memory in Inngest AgentKit 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