Max

What Max does

Max syncs data from external APIs (GitHub, Linear, Google Drive, HubSpot) into local SQLite through typed connectors. Queries run locally - millisecond latency, no rate limits, no network round-trips. Designed so AI agents have fast, unfettered data access.

Use Max

# create a workspace
max init .

# add your favourite connectors
max install --collection git@github.com:max-hq/max-connectors.git
# > added 8 connectors

# connect a couple of sources
max connect @max/connector-github --name pi-mono-repo
max connect @max/connector-linear
max connect @max/connector-claude-code-conversations --name my-chats

# sync a connector
max sync pi-mono-repo
# Syncing...
#     GitHubRepo       █████      12  1021.8 op/s
#     GitHubUser       █████     283  4391.1 op/s
#     GitHubIssue      ███▒·    2156  4811.3 op/s
#     ──────────────────────────────────────────────
#     3.2s elapsed

# query it
max search pi-mono-repo GitHubIssue \
  --filter="state = open AND labels ~= bug" \
  --fields="title,author"

# tell your agent about max, so it's connected to your data
# > Hey Claude - run `max -g llm-bootstrap`

Write a Connector

// Define your entities
const GitHubIssue = EntityDef.create("GitHubIssue", {
  title: Field.string(),
  state: Field.string(),
  author: Field.ref(GitHubUser),
});

// Create a loader that fetches from your API
const IssueLoader = Loader.entity({
  name: "github:issue:basic",
  context: GitHubContext,
  entity: GitHubIssue,
  async load(ref, env) {
    const issue = await env.ops.execute(GetIssue, { id: ref.id });
    return EntityInput.create(ref, {
      title: issue.title,
      state: issue.state,
      author: GitHubUser.ref(issue.author.id),
    });
  },
});

// Wire fields to loaders
const IssueResolver = Resolver.for(GitHubIssue, {
  title: IssueLoader.field("title"),
  state: IssueLoader.field("state"),
  author: IssueLoader.field("author"),
});

// Package it up
export default ConnectorModule.create({
  def: ConnectorDef.create({
    name: "github",
    schema: GitHubSchema,
    resolvers: [IssueResolver, UserResolver, RepoResolver],
    seeder: GitHubSeeder,
    onboarding: GitHubOnboarding,
    operations: [...GitHubOperations],
  }),
  initialise(config, credentials) {
    return Installation.create({ context, start, stop, health });
  },
});

The connector SDK tutorial walks through this step by step.

Embed Max

import { GlobalMax } from "@max/federation";
import { BunPlatform } from "@max/platform-bun";

// Boot the platform
const max = await BunPlatform.createGlobalMax();

// Open a workspace and installation
const workspace = await max.workspace("./my-project");
const installation = await max.installation("pi-mono-repo");

// Sync
const handle = await installation.sync();
await handle.completion();

// Query - local SQLite, millisecond responses
const issues = await installation.engine.query(
  Query.from(GitHubIssue)
    .where("state", "=", "open")
    .where("labels", "~=", "bug")
    .select("title", "author")
);

Max runs as a library. Same engine, same connectors - no CLI required.

Get started

New to Max? The getting started guide walks through creating a workspace, connecting a source, and running your first sync.

Write a connector

The connector SDK tutorial builds a complete connector in six parts - from defining entities to publishing an installable package. The SDK reference covers each subsystem in detail.

Architecture

How Max is structured and why - package layers, federation model, and module boundaries.