Agentic data workflows on the Weld API

Give your agents and scripts the whole data stack. Trigger syncs, publish models, run orchestrations and activate data through the Weld Connect API.

Start for free Read the docs

OpenAPI spec
One key per workspace
SOC 2 Type II

Ask an agent

APIhttps://connect.weld.app · x-api-key

# Board meeting at 9. Refresh revenue and push it to sales.

POST /elt_streams/{id}/request_run        shopify.orders
POST /elt_streams/{id}/request_run        stripe.charges
POST /transforms/{id}/publish             revenue_by_channel
POST /orchestrations/{id}/request_run     daily_revenue
POST /reverse_etl_syncs/{id}/request_run  hubspot.companies

✓ Revenue refreshed, modeled and synced to HubSpot.
▌

Trusted by data and AI teams

Features

Everything in the app. Callable by an agent.

The same syncs, models, orchestrations and destinations you manage by hand, reached through one REST API.

An endpoint for everything

Connections, syncs, streams, models, orchestrations and reverse ETL are all in the Weld Connect API, so an agent can do what you do in the app.

serverhttps://connect.weld.app

Orchestration on demand

Create orchestrations and request runs over the API, or have Weld trigger your dbt Cloud jobs through webhooks when a sync finishes.

POST/orchestrations/{id}/request_run

Your orchestrator, too

Call the API from Dagster, Airflow or Prefect to start and watch pipelines across ingestion, transformation and activation.

dagweld_daily · refresh → model → activate

Models an agent can ship

Agents preview a query against the warehouse, then create and publish it as a model through the Transforms API.

POST/transforms/query_previews

Activation by agents

Agents create, activate and run reverse ETL syncs to push modeled data into CRMs, ad platforms and spreadsheets.

POST/reverse_etl_syncs/{id}/activate

Chat agents over MCP

For Claude, ChatGPT or Cursor, the Weld MCP server puts the same actions behind plain-language prompts.

mcphttps://connect.weld.app/mcp

From your orchestrator

A few lines of Python. The whole pipeline runs.

Airflow, Dagster, Prefect or a cron job: anything that can send an HTTP request can run Weld, and wait for it to finish.

PYdags/weld_daily.py

# Refresh, model and activate, from your own scheduler
import requests

BASE = "https://connect.weld.app"
H = {"x-api-key": WELD_API_KEY}

run = requests.post(f"{BASE}/orchestrations/{ORCH_ID}/request_run", headers=H).json()
status = requests.get(f"{BASE}/orchestration_runs/{run['id']}", headers=H).json()["status"]
▌

endpoints in the Weld Connect API

tools on the Weld MCP server

connectors an agent can set up

Works with

Bring your own agent. It reaches the warehouse through Weld.

Agents provision connectors, ship models and push data back out, while your credentials stay in Weld and your data stays in your warehouse.

Agents and editors

Claude
Claude Code
ChatGPT
Cursor
Copilot

Weld Connect APIand the Weld MCP server

What they can run

dbt
Cube
BigQuery
Snowflake
HubSpot
Salesforce

Explore the Weld Connect API

Customer story

“Weld is just the first building block in our AI vision. We need to ingest the data, we need a single source of truth, we need metrics defined, and in order to do that, we need a data platform.”