August 2026
A redesigned, reflective node I/O model and Configure with AI across many more nodes, plus app event triggers, per-agent models, in-editor hints, the FluxAgent browser agent, and Companion improvements.
This month focuses on making workflows faster to build and easier to reason about. We rolled out a redesigned Input/Output model across the workflow builder — your node's I/O now reflects its configuration automatically — and expanded Configure with AI to many more nodes so you can describe an outcome and let FlowGenX fill in the details. Alongside that: start workflows from app events, pick a model per agent, get guided hints as you build, automate the browser with the new FluxAgent, and do more without leaving Companion.
A simpler, reflective Input & Output model
You no longer maintain a separate I/O schema by hand. Configure a node the normal way and its Input now mirrors that configuration, while the Output shape is supplied by the platform. The redesign is live across the data and control-flow nodes — Data Ingestor, Data Ingestor Advanced, Read File, Write File, Write to Database, SFTP / File Transfer, Set Data, Router, Loop, Knowledge Base, and the App Connector — for a consistent, less error-prone building experience.
Describe the outcome, let AI configure the node
Configure with AI is now available on many more nodes. Describe what you want the node to do and FlowGenX proposes the settings — connection, object, mappings, and conditions — which you review and apply. This release adds the button to the App Connector, Data Ingestor, Data Ingestor Advanced, Read File, Write File, Write to Database, SFTP Transfer, Set Data, Router, Loop, and Knowledge Base nodes.
Guided next steps after applying a proposal
After you apply an AI proposal on the Write to Database node, the assistant now stays open with an "Applied — what to do next" panel so you know exactly what to finish. The selected-object picker also reflects the applied connection immediately, so the schema and table fill in as soon as you connect instead of appearing blank.
Start workflows from app events
Kick off a workflow automatically when something happens in a connected app — a new Slack or Google Chat message, and more. Set up the trigger with Configure with AI, and manage its endpoint lifecycle (enable, disable, and revive) right from the node.
Choose a model per agent
Assign a different LLM to each agent in the designer, so you can match the model to the job — a fast model for routing, a stronger one for reasoning — without splitting your workflow apart.
Guided hints while you build
Contextual, node-by-node hints point out what a node needs and what to do next, so you spend less time guessing at configuration and more time shipping the workflow.
FluxAgent — browser automation
A new agent that drives a real browser to complete web tasks end-to-end, with human-in-the-loop checkpoints for the steps that matter. Test it right on the canvas before you ship.
Companion gets more hands-on
Connect apps without leaving the chat — Companion opens the real connection drawer, resumes an unfinished setup where you left off, and suggests apps when a connector isn't found. A new Settings → Tools tab lets you see and disable the tools the assistant can use.
Jump from logs straight to a run's flow
Open a run's Execution Flow directly from the Logs service graph, and view a workflow's runs right from its trigger — fewer clicks between spotting an issue and seeing exactly what happened.
- Companion: handles apps with no connector gracefully, keeps the workflow preview so “Open in canvas” survives a re-save, shows your messages on reload, and reflects new conversations and auto-naming in the list.
- Knowledge Base: reflects its configuration in Input, sends the selected collections, picks AI-proposed collections reliably, and resolves the query expression in Test Search.
- MCP: External Stdio servers connect via command + args and save with the correct server type.
- Tracing & Logs: traces and logs are scoped to the current tenant, with consistent ordering, reliable first load, empty states for log-less traces, and no polling that resurrects after a modal closes.
- Templates: reliable loading with a graph preview and duplicate-name guards, plus cleaner gallery cards.
- Coworker: better memory, semantic memory and knowledge search, steadier skill learning, and more reliable tool handling with guardrails.
- Webhook triggers: stable, immutable endpoint URLs, stricter timeout parsing, correct sync on Save (not just Test), per-node config so late-arriving ids can't discard your edits, and the ability to revive a disabled endpoint.
- Triggers: one API record per workflow and app type, and app-event icons fall back to a local logo when a signed image URL fails.
- App Connector: applies AI-proposed field values reliably and uses the selected connection's own auth config.
- Stability: resolved several SSR / white-screen crashes (workflow editor, JSON editor, importing a redacted workflow) and quieted noisy toasts on polling endpoints.
- Data & integration nodes: SFTP keeps its upstream Input mappings across save/reload; Read & Write File detect file outputs when binding upstream and clear stale inputs on re-apply; Set Data keeps AI-applied mappings across schema-mode changes; and Configure with AI connection lists are scoped to the selected type and load reliably on large tenants.
- Workflows: new blank workflows get a unique default name automatically.
See it in action
Request a Demo