Self-hosted, controlled use
Hosted on your infrastructure. Multi-account isolation and secret masking are not at sale quality: do not expose the instance to several tenants.
WorkflowBuilder ProPrototype — not for sale
Self-hosted automation, on an audited core scope
A webhook arrives, your data is transformed, a service is called, a response goes out: WorkflowBuilder Pro chains those steps in a visual editor, on your infrastructure, with per-node history. The core covers webhook, HTTP, transform, and SQL. An n8n JSON can be imported for that scope; any missing node is rejected.
The screens shown are real product screenshots. Visible data is demonstration fixture data.
Status
Prototype — not for sale
Automated tests
Sep 2: 12,865 passing
Catalogue
32 visible / 988 types
Audit
4 core scenarios · Sep 2
What it does
A useful function, and a behaviour you can actually inspect.
Triggers
Synchronous webhook, interval scheduler, and file upload have an execution proof. More than forty other triggers are listed, but not audited.
Nodes
The core scope covers data transform, code, HTTP request, SQL, file write, and PostgreSQL.
Import an n8n workflow
An n8n JSON can be imported for the core scope (webhook, HTTP, transform, SQL). Any missing node is rejected; Stripe and Gmail triggers are not armed.
Ten canonical scenarios
Ten scenarios run against deterministic local stubs. That does not prove HMAC anti-replay or a reliable business retry.
Persistent per-node history
Persistent per-node history: status, duration, and output. Replay is not yet a safe contract: it can restart the run from scratch.
Work queue and PostgreSQL
Work queue (BullMQ, Redis) for the scheduler; PostgreSQL database.
Real screenshots
The product as of September 2, 2026.
Every screen below comes from a real run. Visible limitations are left in the presentation.
02
A published, active visual workflow
A webhook arrives, your data is transformed, a service is called, a response goes out: WorkflowBuilder Pro chains those steps in a visual editor, on your infrastructure, with per-node history.
03
Follow executions
Persistent per-node history: status, duration, and output. Replay is not yet a safe contract.
Why Agile Up
Simple engineering choices, before they become marketing claims.
Evidence available
Four core scenarios have external evidence. The engine can still announce success with no real effect for some missing nodes, triggers, or retries.
No mandatory subscription
Self-hosted, with no mandatory SaaS rent. Code under Business Source License 1.1 (BUSL-1.1), not MIT.
Honest limits
What still needs to be done or verified.
- Catalogue: 988 types. The editor shows 32 by default; the other 956 are experimental and hidden. No integration with a Slack, Gmail, Stripe, or other external account is guaranteed. Licence: BUSL-1.1.
- Twelve engine defects were reproduced on September 3 (isolation, secrets, batched HTTP, queue, n8n, HMAC, retry, cron, merge, and others); they are being fixed. Local patches are not yet merged.
- History is persistent, but replay can restart the run from scratch, and isolation between accounts is not ready.
- No public instance and no online payment: demo by appointment.
Product status
An audited prototype, not for sale as it stands.
As of September 2, 2026, the existing suite had 12,865 passing tests. On September 3, 12 engine defects were reproduced — tenant isolation, secret leakage, batched HTTP, queue, n8n import, HMAC, retry, cron, and merge among them — and are being fixed. On September 2, four core scenarios also ran end to end with effects verified outside the application.
- Ten scenarios run against deterministic local stubs; that is not evidence of production robustness.
- Audited core scope: webhook, HTTP, transform, SQL, and files. A missing node is rejected.
- The 12 defects reproduced on September 3 are being fixed; local patches are not merged here.
- No public instance and no online payment: demo by appointment.
Agile Up · Software publishing
A question about WorkflowBuilder Pro or its current scope?