@glidecms/connect-sdk

TypeScript SDK for the Glide Connect content delivery API. Read-only access to published articles, taxonomies, collections, images, and all CMS content types.

Zero runtime dependencies. Native fetch. Typed errors. Optional LRU caching. Schema auto-discovery.

Install

npm install @glidecms/connect-sdk

Requires Node.js >= 22.

Quick start

import { Client } from '@glidecms/connect-sdk';

const sdk = new Client({
  apiUrl: 'https://connect.your-tenant.gcpp.io/v1',
  apiKey: 'your-api-key',
});

const articles = await sdk.article.list();
const taxonomy = await sdk.taxonomy.getById(42);

API surface

26 clients — all flat on the sdk object:

accessBundle, article, author, articleType, changeHistory, collection, collectionType, contentPanel, contentTag, customFieldGroup, file, gallery, glideEntity, image, liveReport, liveReportType, page, redirect, route, settings, siteConfig, taxonomy, taxonomyConfiguration, template, widgetData, widgetType

Plus sdk.schema for auto-discovery of article types, taxonomies, authors, collection types, and custom field groups.

Client methods

Every client extends BaseClient<T> and provides:

| Method | Returns | Description | |---|---|---| | getById(id) | Promise<T> | Fetch a single item by ID | | getByIds(ids) | AsyncGenerator<T[]> | Batch fetch by IDs (yields batches of 50) | | list(opts?) | Promise<T[]> | Fetch items matching query | | listWithMeta(opts?) | Promise<ListResult<T>> | Fetch items with total count and metadata | | listAll(opts?) | AsyncGenerator<T[]> | Iterate all items across all pages | | aggregate(opts) | Promise<AggregateResult<T>> | Run aggregation queries | | paginate(opts?, pageSize?) | Paginator<T> | Manual page-by-page iteration | | waitFor(id, opts?) | Promise<T> | Poll until item exists and optional condition passes |

Some clients add resource-specific methods (e.g. ArticleClient.search(), LiveReportClient.getPosts()).

AI-assisted usage

For SDK selection, safe defaults, task-oriented method guidance, and mutation guardrails, see Agent guide.

Documentation

| | | |---|---| | Tutorial | Build your first Connect integration | | How-to guides | Filtering, search, pagination, caching, logging, query builder, schema, error handling, retry, waitFor, utilities | | API reference | Full method signatures for all clients, config, errors | | Migration notes | Async iteration and request option shape changes | | Agent guide | Agent/MCP usage guidance for read-only delivery workflows | | Agent readiness | Field resolution, query validation, capabilities, content helpers for MCP/CLI/agents |

Guarantees

• Request options. signal, timeout, retry, and cache are honored by getById, list, listAll, fetchAll, paginate, aggregate, and waitFor.
• Timeouts. Default 30s. Per-request override via opts.timeout.
• Cancellation. All methods accept AbortSignal. Aborted requests throw immediately without retry.
• Retry. 5xx and network errors retried with exponential backoff. 4xx, aborts, and timeouts are never retried.
• Rate limiting. 429 responses throw ConnectRateLimitError with retryAfter populated from the response.
• waitFor(). Polls with configurable timeout and interval. Each poll is capped to remaining time. Accepts AbortSignal. Timeout error distinguishes "not found" from "condition not met".
• listAll(). Signal, timeout, retry, and cache propagate through all internal pagination and probe requests, including date-window splitting for large datasets.
• Deduplication. Concurrent identical GETs without per-request opts share one fetch. GETs with opts always make independent requests.