Monthly Dispatch: October 2026

If you have built a workflow in Workato or n8n, you know the shape: boxes on a canvas, each one a step, wired together into a pipeline. Apache NiFi is the same idea built for heavier data work. This client moved from Workato to NiFi within the past year because NiFi handled more sophisticated flows, and their installation grew to roughly two thousand of those boxes, which NiFi calls processors: feeds in, transformations, lookups, and loads into the warehouse. Every change meant opening the canvas, tracing the flow by hand, and hoping nothing downstream depended on the piece you were about to touch.

Canvas tools like these exist because of a tradeoff. Custom software could do anything, but it was expensive to build and harder still to develop, test, and maintain properly, so teams accepted a canvas and its limits instead. An agent working inside the right guardrails removes that tradeoff. On Plexara you get the power of custom code without paying developers to write it or to keep it running: the agent writes it, every version is tested against real runs before it can be saved, and you approve any change in what it does. That code also works natively with your entire data estate: the warehouse, the catalog and its business definitions, your files, and every API connection your team has set up, under the same access controls as everything else.

That work now runs as Plexara automations. The agent read each flow, wrote the script that does the same job in Starlark, a small, readable scripting language, and ran it beside the old flow. A flow was switched off only after a recorded comparison showed the new output matched the old rows. One flow at a time, the canvas emptied out.

The difference shows up after cutover. Each automation is a short script with a version history and a record of every run: what triggered it, what it read, what it wrote, and how it ended. A failure arrives as an email with the reason and the log. Maintenance is a conversation: "the settlement load ran long last night, find out why and fix it." The agent reads the run record, proposes the change, and tests it against previous runs. If the fix would change what the automation does, you get a plain-language summary and decide whether it goes ahead. Nobody has to read code to approve it.

See Automations, and Plexara compared with pipeline tools for how this differs from putting an agent next to a flow product.

Describe a process and the agent builds the automation, tests it against real results, and checks every new version against the runs before it. When a change would make it behave differently, you see what changes in plain language, approve or decline, and the approval is kept with the version.

Automations are now shared across your team. The portal section is called Automations, everyone can see the automations that run for them and how each run went, and an owner can let a colleague, a role, or an application run a finished automation from a simple form with live progress. A diagram of each automation shows what it reads, writes, and sends.

Your CRM, email platform, point-of-sale system, or any tool that sends webhooks can now send its events straight to Plexara. Every event is kept and is queryable alongside the rest of your data the moment it arrives, and automations turn those events into warehouse updates, reports, and alerts. See Webhooks.

Automations can post to a Mattermost channel, an incoming webhook, or a named email list. Your admins set up the channels in the portal and confirm each with a test send. Automations can also deliver results to other systems: each delivery is signed, so the receiver knows it came from your Plexara, and a retried delivery is never processed twice. See Notifications.

Plexara now connects to Tableau, MicroStrategy, and Veeva Vault, and keeps those connections signed in so work never stops for an expired session. GraphQL and SOAP services connect the same way REST APIs do: the agent finds the operation it needs by describing what it wants. Every connection has its own page under API Connections.

Ask the agent to save a map to Plexara and it builds one as a report in the portal: revenue by state, members by county, stores and deliveries on a street map. The map engine and the street basemap are served from your own Plexara deployment, so opening a map, sharing it by link, or exporting it to PDF never sends a request to an outside map service. Nothing about what you mapped, or who looked at it, leaves your deployment.

Street maps cover the regions your admins choose in the portal under Admin settings: the contiguous United States, Alaska, Hawaii, Puerto Rico, Canada, Mexico, the United Kingdom and Ireland, Europe, Australia, or a custom area. The settings page shows a region's exact size before it is saved and its progress while it is prepared. The agent knows which regions you have, and maps by state or county need no region at all.

Ask the agent to save a deck to Plexara and the portal presents it as slides, with export to PDF. Say “save it to Plexara” in the request: left to itself, an assistant will usually build the file on your own machine instead. Reports export to PDF exactly as they look on screen, and automations produce formatted Excel workbooks. Going the other direction, PDFs, Word and Excel files, decks, and images in your library now reach the agent as their contents, so it can read what you hand it. Feedback on your work and notifications now share one Inbox.

A saved query result or automation output updates the same file on every refresh, and becomes a queryable table in the same step, so the reports built on it keep working.

Full notes are on the changelog.