Products

Solutions

More

Data agents Beta

Where data agents
are built and scaled.

Factory is the platform underneath Radar and Benchmark. Describe the data you need in plain English — an agent builds the pipeline, runs it on a schedule, and rebuilds it when the web changes.

Just want the data, not the machinery? See Radar and Benchmark.

factory.jsonify.com

Home / Build

3
Understand

Ask questions and explore your workspace data.

Build

Create pipelines and monitor their runs.

Global Market Radar

What do you want to build?

Build a pipeline, analyze a file, or ask anything…
Attach

Pipelines

5
Change detection Last run 21h ago
Inventory monitoring Last run 23h ago
Price & promotion tracking Last run 1d ago
Product identity matching Last run 1d ago
Catalogue discovery Last run 1d ago
View all

Datasets

5
Live product offers8,426,391 rows • Updated 5m ago
Price observations6,843,120 rows • Updated 20m ago
Inventory signals5,206,774 rows • Updated 35m ago
Product matches4,982,771 rows • Updated 50m ago
Daily market rollups1,274,880 rows • Updated 1h ago
View all

Recent runs

Change detection184,070 rows • 21h ago
Inventory monitoring644,800 rows • 23h ago
Price & promotion tracking714,519 rows • 1d ago
Product identity matching498,410 rows • 1d ago
Catalogue discovery211,676 rows • 1d ago
View all

Home / Understand

3
Understand

Ask questions and explore your workspace data.

Build

Create pipelines and monitor their runs.

Global Market Radar

What do you want to know?

Ask a question about your data…
Attach

Sources

12
a
amazon.com1,000+ rows
b
bestbuy.com1,000+ rows
C
costco.com1,000+ rows
e
ebay.com1,000+ rows
H
homedepot.com1,000+ rows
View all

Datasets

5
Live Product Offers8,426,391 rows • Updated 4m ago
Price Observations6,843,120 rows • Updated 19m ago
Inventory Signals5,206,774 rows • Updated 34m ago
Product Matches4,982,771 rows • Updated 49m ago
Daily Market Rollups1,274,880 rows • Updated 1h ago
View all

Recent alerts

3Clear all
18,420 meaningful price drops2h ago • Prices fell by at least 10% across 18,420 tracked offers
42,107 availability changes4h ago • Tracked offers moved into or out of stock
3,284 new products discovered6h ago • New product pages were added to the catalogue
View all

From a sentence to a running pipeline.

No scraping scripts, no selectors, no maintenance rota. You describe the outcome — Factory's agents do the engineering.

You

Hi Jason! I want to collect the prices of all my products from all around the web, every day. I guess probably 100k rows across a dozen sites. I want to collect that, turn it into a dashboard and newsletter for me. Keep it up to date automatically.

Can you help?

Inspect file product_catalog.csv Probe sources 12 retailers Draft schema Live product offers Validate sample 500 rows Estimate volume 8.4M rows / day

Preparing the smallest useful catalog handoff.

Yes — that volume and the twelve-site shape are feasible. I can build the daily collection, canonical product matching, price history, dashboard, alerts, and newsletter.

Took 48s

You

Here's our current product catalog. It has the SKUs, UPCs, priority retailers, and a handful of known product URLs.

Please go ahead, schedule it daily, and fix routine site changes automatically when you can.

Attached: product_catalog.csv

Schedule pipeline Change detection Create dashboard Competitive Market Pulse Save dashboard widget Competitive Market Pulse Save alert Meaningful price drops Test alert Meaningful price drops Newsletter Create Daily Market Brief Newsletter Render Daily Market Brief Run pipeline Price & promotion tracking Query data Price observations

Running final checks and organizing the workspace links.

Task completed

Market Radar is live — 12 retailers, 5 pipelines, 5 datasets, dashboard, newsletter, and alerting are ready.

Open dashboard

Done — I built the Market Radar system from your catalog and verified the first daily run. Here's everything:

  • Live product offers — the canonical 8.4M-row product and offer view
  • Price observations and inventory signals
  • Competitive Market Pulse — the dashboard for coverage and price movement
  • Daily Market Brief — scheduled every morning at 07:00
  • Five daily pipelines with routine site-change recovery and independent validation

Let me know if you need anything else, or if you’d like me to help analyse the data.

Took 1h 3m 42s

Ask Jason to build, inspect, or explain…

Attach
Step 1

Describe the goal

"Track competitor pricing across 40 grocery sites." Plain English is the whole spec — sources, schema, and cadence are inferred or asked for.

Step 2

An agent team builds it

In an isolated sandbox, the agent probes sources, picks fetch strategies, locks a schema, and assembles a multi-step pipeline — then validates it end to end.

Step 3

It runs on schedule

Daily, weekly, or on demand. Datasets, dashboards, and alerts are delivered to Slack, Snowflake, PowerBI, Excel, or your API — and shared via live links.

The machinery under every dataset.

Everything Radar monitors and Benchmark simulates is produced by this stack — built once, then run and repaired automatically.

Agent-built pipelines

An LLM agent plans the extraction, writes the pipeline, and tests it against real pages — the same work a data engineer would do, in minutes.

A fetch chain that learns

10+ backends, from plain HTTP to full browser automation. Factory remembers what works per domain and starts there next time.

Sandboxed execution

Every agent and every pipeline runs in an isolated sandbox — no shared state, no access beyond the job it was given.

Cell-level provenance

Every value in every dataset traces back to the page it came from. Click a cell, see the source — no "trust us" data.

Self-healing runs

When a site redesign breaks a step, a healing agent rebuilds just that step and retries the run — before anyone notices.

Scheduled delivery

Dashboards, alerts, newsletters, and exports — Slack, Snowflake, PowerBI, Excel, CSV, or API — on the cadence you choose.

The web changes. Factory notices first.

Scrapers rot — that's why most teams give up on them. Factory treats a broken step as a job for another agent, not a ticket for your engineers.

pipeline runs / #312 · Grocery Pricing

scheduled 02:00
Ttesco.com
+ 39 more sources
fetch
extract v1v2
heal agent
clean
deliver

No ticket raised. No engineer paged. The morning dataset arrived on schedule.

Every factory needs a foreman.

Jason is the copilot inside every Factory workspace. Ask him anything about your data and he answers from the live datasets — then acts on what you decide, right in the chat.

  • Answers from your live data

    "Why did prices drop yesterday?" gets a real answer, queried from the dataset — not a canned reply.

  • Builds and edits pipelines

    "Track two more retailers" or "add a promo column" — Jason changes the pipeline, validates it with a sample build, and versions every edit so you can revert.

  • Runs the delivery side too

    Dashboard widgets, alerts, newsletters, scheduled summaries — and when a build or rerun finishes, he reports back in the same thread, in the app or in Slack.

J

Jason

watching your workspace

web · slack · cli
Why did Tesco prices drop yesterday?
query_data · price_history · 1,904 rows
12 spirits SKUs dropped an average of 8% — a category-wide promotion that started Tuesday. Want me to alert you when it ends?
Yes — and send me a weekly summary.
schedule_message · weekly · mon 08:00
Done. I'll watch the promotion and post a summary here every Monday at 08:00. ✓
Track Aldi and Lidl too — same columns.
edit_pipeline · +2 sources · sample build started
On it. I'm adding both retailers to the fetch step and reusing your schema — running a sample build to validate first. I'll report back here.
run #341 finished ✓ · 96 sample rows
Build finished — Aldi and Lidl are flowing in clean. Saved as v2 and added to tomorrow's 06:00 run. You can revert to v1 anytime.
Put our price gap vs competitors on the dashboard.
render_chart · price_gap_by_retailer
Added a "Price gap vs market" widget to your dashboard. Want it in Monday's newsletter too?
Yes — and alert me if any gap passes 5%.
create_alert · gap > 5% → slack
Done. The widget is in the newsletter draft, and I'll ping you in Slack the moment a SKU crosses 5%. ✓

Ask Jason about your data…

answers · builds · delivers

Inspectable by design.

You don't have to look inside — but you always can. Every pipeline Factory builds is real, versioned, testable software.

Code is the source of truth

Factory generates Python pipelines — not opaque configuration. Every pipeline is readable, diffable, and reviewable, like any other code your team ships.

Skills: extraction knowledge, versioned

Extraction logic is packaged as skills — versioned, installable units with typed input/output schemas, tests, and fixtures, stored in a registry. Skills are domain-aware and reusable, so every pipeline in a domain gets smarter than the last.

Every pipeline is a DAG

Pipelines render as inspectable DAGs — sources, extraction steps, transforms, and outputs, visible end to end. When something breaks, you can see exactly where.

Typed schemas, validated every run

Outputs are typed and validated against their schema on every run. Anomalies get flagged, accuracy gets tracked — 98.7% extraction accuracy across complex web & app sources.

Sandboxed execution, escalating fetch

Pipelines execute in isolated sandboxes. The fetch layer escalates automatically — plain HTML, headless browser, and beyond, across 10+ backends — until the data flows.

One factory, two products

Every Radar dataset and every Benchmark table is a Factory pipeline underneath. Most teams meet Factory through one of them.

Who Factory is built for.

Built for people who read the code before they trust the output.

Engineers who've maintained scrapers

You've written the XPath, watched it break, and rewritten it. Factory generates the pipeline, watches the source, and repairs itself — you review DAGs and diffs instead of firefighting.

Technical evaluators

Your strategy team is already talking to us. This page is your diligence stop: generated code, typed schemas, versioned skills, sandboxed execution, DAG-level observability. Ask us anything — or book a call and we'll walk the architecture.

Platform teams building on structured web data

You need external data as a dependable input, not a side project. Typed, validated, scheduled outputs delivered into your warehouse — with the pipeline itself inspectable when you need it.

Run it yourself, or let us run it for you.

Demo access

Free

Build pipelines and run them on synthetic demo data. Free to explore — no card required.

Self-serve

Usage-based

Pay per pipeline run and rows extracted. You build, you operate, you export. Scales with what you actually use.

Managed

Custom

We build, run, monitor, and guarantee your pipelines. SLAs, quality validation on every run, dedicated infrastructure, delivery straight into your stack.

Book a call →

Self-serve tiers cover pipelines you operate yourself. Managed datasets — built, monitored, and quality-guaranteed by Jsonify — are scoped per engagement.

SOC 2 Type 2 · GDPR compliant · Independent controller model — see legal FAQ

FAQ

What about personal data and GDPR?
Jsonify operates as an independent data controller. Deliverables are structured and aggregated — customers never receive raw personal data, so there's typically no DPA required and nothing sensitive lands in your systems. We're SOC 2 Type 2 audited and GDPR compliant. Details in the legal FAQ.
Can I see what a pipeline actually does?
Yes. Every pipeline renders as a DAG and is backed by generated Python you can read. Skills carry typed schemas, tests, and fixtures. Nothing load-bearing is hidden.

Watch Factory build your first pipeline.

Describe the data you need and see a pipeline assembled live — runs instantly with demo data, no setup required.