May 2026

New connectors

Linear

You can now sync data from Linear — the project management tool for software teams. The initial release includes comments, issues, teams, and users as ELT source streams, plus Reverse ETL support for writing data back to Linear.

Explore Linear documentation →

Linear Reverse ETL

Linear now supports Reverse ETL with insert, update, and upsert operations. Push data from your SQL models, dbt models, or raw warehouse tables back into Linear to create or update issues and comments programmatically.

Amazon Vendor Central

Amazon Vendor Central is a new connector for manufacturers and distributors selling directly to Amazon. It syncs vendor-specific sales, inventory, traffic, forecasting, and margin reports — with both daily and real-time variants.

Explore Amazon Vendor Central documentation →

New features

MCP Server

Weld now exposes an MCP (Model Context Protocol) server, allowing AI coding tools like Claude Desktop, Cursor, and Windsurf to interact with your Weld workspace directly. The MCP server supports the same operations as the REST API — managing connections, ELT syncs, streams, transforms, and orchestrations — plus additional debugging tools like sync history inspection and failure listing.

Authentication works via API key or OAuth, so AI agents can query your data pipeline status, trigger syncs, and preview transform queries conversationally.

Weld Connect API: Orchestration Endpoints

The REST API now includes full Orchestration support:

• GET /orchestrations — list all orchestrations (paginated)
• GET /orchestrations/:id — get a specific orchestration
• POST /orchestrations/:id/request_run — trigger an immediate run
• GET /orchestration_runs — list runs (with optional orchestration filter)
• GET /orchestration_runs/:id — get a specific run

Weld Connect API: ELT Stream Runs

You can now inspect ELT stream run history via the REST API:

• GET /elt_stream_runs — list runs for a stream (paginated, sorted by start time)
• GET /elt_stream_runs/:id — get a specific run by ID
• GET /elt_streams/:id — get stream details

The request_run endpoint now also returns a status_url for polling and includes the running_job_id on conflict responses.

GitHub Integration for Data Models

You can now connect a GitHub repository to manage your data models as code. Once connected, model definitions live in your repo and Weld enforces read-only mode in the UI — preventing accidental edits to code-managed models.

The setup wizard lets you pick a repository and branch from your GitHub App installation. If the installation is later removed, Weld automatically marks the connection as broken and notifies you.

Sync History Redesign

The data source History tab has been redesigned with a two-level UI:

• All-tables view — shows latest status per source stream with a visual timeline
• Detail view — per-stream paginated history with a clickable timeline chart, status filters, and "hide empty" toggle

The new timeline chart is fully interactive: click a bar to jump to that run, hover for details, and use resolution controls (hour / day / week). History is now cursor-paginated so you can scroll back through the full run history.

Stream Deprecation Notifications

Weld now sends in-app notifications and emails when a synced table or subtable is deprecated by the source connector. Notifications include the deprecation date, replacement table (if available), and a direct link to your sync configuration. This gives you advance warning to migrate before a table stops syncing.

Substream Exclusion

You can now exclude individual substreams from syncing when configuring a connector. Previously, enabling a main stream would always sync all of its substreams. Now you can deselect specific substreams you don't need — reducing warehouse clutter and sync time.

Deprecated substreams are automatically excluded for newly added streams and hidden from the UI when not selected.

REST API: Materialization Status & Wait for Completion

The REST API for data models now returns materialization_status and materialization_error fields on transform responses. You can also pass wait_for_completion: true when creating, updating, or publishing a model to block until table materialization finishes — useful for CI/CD pipelines that need to confirm a model is fully available before proceeding.

BigQuery Connection Editing

You can now update your BigQuery data warehouse connection credentials directly from Settings, without needing to re-create the connection. Project ID and location are validated to prevent accidental changes.

Connect API on Premium Plan

The Weld Connect API is now available on the Premium plan, giving more teams programmatic access to manage syncs, orchestrations, and data models.

Per-Stream Custom Schedules on Orchestrated Syncs

Individual source streams can now opt out of orchestration-managed scheduling by setting their own sync interval. A stream with a custom schedule runs independently while the rest of the source continues to follow the orchestration workflow. Clearing the custom interval re-enrolls the stream in orchestration.

The schedule configuration UI now also shows a warning when changing a stream's schedule will override orchestration timing, and you can set a custom schedule on tables that are still running their initial sync.

MCP Server: Role-Based Access

The MCP server now respects workspace roles. Tools are only offered to AI agents based on the authenticated user's permissions — Viewers see read-only tools, Editors see read + write tools, and users with no role see nothing. This prevents AI agents from performing actions the user wouldn't be allowed to perform in the UI.

Connector updates

Amazon Selling Partner & Vendor Central: Authentication Choice Restored

The Amazon Selling Partner and Amazon Vendor Central connectors now support both authentication methods side by side. You can choose between:

• Weld's shared OAuth — the original method, no SP-API app registration needed
• Private app credentials — authenticate with your own Refresh Token, Client ID, and Client Secret

Shopify Reverse ETL

Shopify now supports Reverse ETL with Upsert mode for two object types:

• customer — upsert customers with email, phone, name, addresses, tags, tax settings, and locale.
• metafield — bulk set metafields on any Shopify resource (owner ID, namespace, key, type, and value).

Explore Shopify Reverse ETL setup guide →

Google Analytics 4: Date Dimension Selection

GA4 custom reports now let you select a date dimension granularity — daily, weekly, monthly, yearly, or ISO variants. Previously all reports used the date dimension. The selected granularity controls how report periods are sliced, and each row now includes explicit date_range_start and date_range_end columns.

Microsoft Ads: Multi-Account Resilience

Microsoft Ads syncs no longer fail entirely when access to one ad account is lost. The affected account is skipped with a notification, and syncing continues for all remaining accounts. A new MICROSOFT_ADS_ACCOUNT_ACCESS_LOST notification alerts you to re-authorize the affected account.

HubSpot: Associations Pagination

HubSpot v4 association batch reads now correctly paginate when an object has more associations than fit in a single API response. Previously, only the first page of associations was returned.

HubSpot: Association Tables — Known Limitation

HubSpot: Submissions Primary Key Fix

Added object_type_id to the HubSpot submission_value substream primary key. Previously, name alone wasn't unique — form fields with the same name but different object types caused duplicate key collisions.

Klaviyo: Filtered Event Tables

Klaviyo now supports metric-based filtered event tables. You can select specific Klaviyo metrics (e.g. "Placed Order", "Opened Email") and Weld will create dedicated event tables filtered to those metrics — giving you smaller, faster-syncing tables focused on the events you care about.

Amazon Ads: Sponsored Products v3 Migration

Completed the migration of remaining Amazon Ads Sponsored Products endpoints to v3: negative keywords, campaign negative keywords, targeting, negative targeting, and portfolios now use the latest Amazon Ads API.

Google Analytics 4

Added the sessionPrimaryChannelGroup dimension to GA4.

Cargoson

Added two new tables to Cargoson:

Facebook Ads: Graceful Handling of Inaccessible Ad Accounts

Facebook Ads syncs no longer fail entirely when one ad account becomes inaccessible. Inaccessible accounts are now skipped with a warning, and the sync continues for all remaining accessible accounts.

Facebook Ads: Incremental Creatives

The Facebook Ads creatives stream now supports incremental syncing, avoiding full-refresh behavior on each sync.

Facebook Ads: API v24 Upgrade

Upgraded the Facebook Ads SDK to API version 24, matching the already-configured API version.

Amazon Seller Central: Catalog Items Stream

Amazon Seller Central now includes a catalog_item stream with substreams for product attributes, dimensions, identifiers, images, relationships, and sales rankings.

Amazon Vendor Central: Purchase Orders & Direct Fulfillment Orders

Amazon Vendor Central now includes purchase orders and direct fulfillment orders as source streams, with item-level substreams for both.

Google Analytics 4: Configurable Backfill Days

GA4 custom reports now let you configure which trailing backfill days to include (e.g. 1, 7, 30 days). Previously, backfill windows were fixed. You can now select specific day offsets (2–29) to control how far back each sync re-fetches data for completeness.

Google Analytics 4: Pagination Fix

GA4 report data is now fully paginated. Previously, only the first page of results was returned for large reports. All rows are now fetched using offset-based pagination.

Google Ads: 37-Month Lookback Limit

Google Ads now clamps incremental sync start dates to a maximum of 37 months in the past, matching Google's updated granular data retention window. This prevents DateRangeError.INVALID_DATE errors that began occurring in June 2026.

Shopify: ShopifyQL Custom Reports

Shopify now supports custom reports powered by ShopifyQL — Shopify's commerce-specific query language. This works like Shopify's built-in Reports module but syncs the results directly into your data warehouse as a table.

You can create up to 20 custom reports per connection. Each report validates your ShopifyQL query, previews the output columns and types, and supports incremental sync so only new data is fetched on subsequent runs.

Example — sync daily session metrics filtered to human traffic:

FROM sessions
SHOW online_store_visitors, sessions, average_session_duration,
added_to_cart_rate, reached_checkout_rate, conversion_rate, bounce_rate
WHERE human_or_bot_session IN ('human')
GROUP BY day WITH TOTALS
TIMESERIES day

Read the full Custom Reports guide →

Shopify: Access Token Authentication

Shopify with Access Token now supports connecting via a user-provided access token as an alternative to the standard OAuth flow. This is useful for custom apps or scenarios where OAuth isn't available.

Gorgias: Customer Stream

A new customer stream is now available for Gorgias, syncing customer records with incremental support based on updated_datetime.

Gorgias: Ticket-Customer-Updated Events

Gorgias ticket syncs now also react to ticket-customer-updated events. Previously, tickets were not re-synced when only their associated customer changed.

TikTok Ads: Big Integer Precision

TikTok Ads responses with numeric IDs longer than 15 characters are now handled correctly using big-integer parsing. Previously, JavaScript's floating-point precision could silently corrupt large numeric IDs.

Linear Reverse ETL: Intercom Attachment

Linear Reverse ETL now supports attaching an Intercom conversation link to a Linear issue. You can include Intercom conversation details when creating or updating issues, or use the dedicated attachment_link_intercom object type.

Traede: Variant Cost Prices

Traede now includes a variant_cost_price substream on the products stream, providing per-variant cost price data including currency, cost price, product costs, total cost, and landed cost.

Salesforce: Column Name Collision Handling

Salesforce now handles cases where multiple field names collapse to the same snake_case representation. Instead of silently dropping duplicate columns, Weld generates unique names with a deterministic suffix so no data is lost.

YouTube Analytics: Stream Lock for Unavailable Report Types

YouTube Analytics now pauses a stream and sends a notification when the configured report type is no longer available from the YouTube Reporting API, instead of failing silently on each sync.

Platform

MySQL CDC: IANA Timezone Selection

You can now select an IANA timezone when configuring MySQL CDC, ensuring timestamps are interpreted correctly for your database's timezone.

PostgreSQL: Read Replica Recovery Conflict Notifications

When a PostgreSQL read replica encounters a recovery conflict during CDC streaming, Weld now sends a dedicated notification with remediation guidance instead of a generic error.

Model Import: Materialization Waiting

When importing models that depend on upstream table materializations, Weld now waits for the upstream materialization to complete before proceeding. This prevents failures caused by downstream models referencing tables that haven't finished materializing yet.

Shopify Sync Resilience

Improved reliability for Shopify syncs: adaptive chunk sizing for large order histories, bulk operation tracking via Redis to prevent conflicts, and smarter backfill start dates based on the shop's actual first order.

ShipHero Sync Resilience

Improved reliability for ShipHero syncs: adaptive page sizes for both parent and sub-queries during complexity throttling, graceful chunk completion on timeouts, and automatic fallback from sorted to unsorted pagination when gateway timeouts persist.

Amazon Seller Central: Order Chunking

Amazon Seller Central order syncs now process data in one-month chunks, improving reliability for accounts with large order volumes.

Amazon SP-API: Exponential Backoff for Report Polling

Report status polling for Amazon Seller Central now uses exponential backoff when the processing status remains unchanged, reducing unnecessary API calls during long-running report generation.

Bug fixes

• Google Sheets: Fixed Google Sheets sync failures when sheet tab names contain colons or other special characters (e.g. input: hourly data).
• Reddit Ads: Fixed a form field mismatch in the Reddit Ads setup form that prevented ad account selection.
• PayPal: Reverted transactions stream from chunked to incremental sync — the API does not reliably sort by transaction_initiation_date, making chunk-based resumption unsafe.
• Traede: Added deduplication to categories, invoices, products, and variants streams to filter out duplicate rows caused by pagination.
• Adyen: Fixed schema errors when CSV column names start with a number (now prefixed with n_).
• Amazon Ads / Attribution: Fixed authentication failures caused by missing region when the region value was stored as a plain string.
• Amazon Selling Partner: Fixed empty reports (status DONE without a document ID) causing infinite polling loops.
• Shopify: Contextual pricing now always fetches the full product variant set — the API doesn't reliably reflect pricing changes in updated_at timestamps.
• TikTok Organic: Fixed pagination for video insights where intermediate pages with no videos would prematurely end the sync.
• Gorgias: Fixed pagination failures when the API returns null data or meta fields.
• Snowflake: Fixed compression handling when downloading large result sets from GCP-backed Snowflake stages.
• Facebook Ads: Reduced excessive retries (from 100 to 10) when the creatives API returns empty pages, preventing unnecessarily long sync times.