Falcon Builder
Deploy production-ready AI agents across your org.
I just launched a self-hosted version of https://FalconBuilder.dev, so you can now run a complete AI orchestration platform on your own machine.
Run:
git clone https://github.com/NeoSky-AI/falcon-builder-self-hosted
cd falcon-builder-self-hosted
./setup.sh # asks for your public URL, generates every secret into .env
docker compose up -d
It's that easy! Take a look at the full repo here: https://github.com/NeoSky-AI/falcon-builder-self-hosted
We got our first enterprise client on Falcon Builder a couple of months ago, bringing our ARR to $18K in pure licensing fees. As we've built out this platform, it has become instrumental for this client, which now has over 600 workflows in production!
-
1600+ production workflows for one enterprise client is where this gets really interesting. At that scale, I've been thinking less about whether an agent/workflow can execute correctly and more about what independently establishes that a consequential action was still authorised at the moment it actually affected another system. Take a workflow that updates a record, sends a customer communication or triggers an external action. It can have valid access, pass its checks, execute the expected path and produce a clean audit trail. But if authority changes between the earlier approval/decision and the downstream consequence, there are really three separate questions: Was it technically permitted? Was it still authorised when the consequence occurred? Can the downstream evidence establish what actually happened? We've been exploring that separation with OpsWatch, particularly because “the workflow says it was blocked” and “we can prove the external consequence did not occur” aren't necessarily the same state. With 600+ workflows now running for the healthcare client, have you encountered cases where Falcon's internal execution evidence wasn't enough by itself and you needed to reconcile against the downstream system to establish the real outcome?
If you've used AI tools for development or automation, you know the drill: copy your code into a chat window, describe what you're working on, explain the context for the tenth time, and hope the AI generates something close enough to be useful. Then you paste it back, fix the parts that don't fit, and repeat.
We built AI Wingman to eliminate that entire loop. Instead of an AI that lives in a separate window and starts from zero every time, Wingman is embedded directly inside the Falcon Builder workflow editor. It sees your nodes, your connections, your configurations, and your data flow. When you ask it to make a change, it doesn't guess — it knows exactly what your workflow looks like, and it proposes targeted modifications that you review before applying.
This post is a technical walkthrough of how we built it, why we made the architectural decisions we did, and what it means for how you build automation.
When I started building Falcon Builder, I kept running into the same problem.
The people who needed automation the most weren't the ones who could build it.
They'd open the editor, see a blank canvas, and freeze.
Not because they weren't smart. But because translating a business problem into nodes, edges, and logic is a skill — and most people don't have time to learn it.
I've been thinking about that gap for a long time.
Today, we're closing it.
AI Wingman is now live inside Falcon Builder — an AI assistant embedded directly in the workflow editor that understands your entire workflow: every node, every connection, every data flow.
Describe what you want to build. It proposes the changes. You review and apply them — one at a time, or all at once.
It's the Cursor experience, but for AI agents and automations.
No more staring at a blank canvas. No more Googling how to wire a condition node. No more rebuilding from scratch because something changed.
Just describe it. Review it. Ship it.
We built this because automation should be accessible to the person with the idea — not just the person who can build it.
If you've been waiting for a reason to try Falcon Builder, this is it.
Check it out 👉 https://falconbuilder.dev
Or read our blog post explaining the details of the build here: https://www.falconbuilder.dev/blog/ai-wingman-workflow-intelligence
Today, we integrated a Duplicate feature for nodes inside of workflows. This will drastically increase the speed of building complex workflows using similar nodes, allowing the user to click Duplicate or use keyboard shortcuts (⌘D on Mac / Ctrl+D on Windows) to duplicate a node with all of its configuration details intact. The duplicate appears near the original, ready to be connected into a new workflow stream.
Today, we launched HighLevel integration into Falcon Builder. Connect your GoHighLevel (HighLevel) CRM to automate contact management, opportunity tracking, and more using OAuth2. The HighLevel node supports creating, reading, updating, and deleting contacts and opportunities, plus a custom API call mode for any HighLevel v2 endpoint. Here is the integration documentation: https://www.falconbuilder.dev/docs/integrations/highlevel
Last week, we had a big feature update - integrating AWS S3 into Falcon Builder. You can now easily manage files in Amazon S3 buckets from your workflows. Upload, download, list, copy, and delete objects, plus create and delete folders. Here is the integration documentation: https://www.falconbuilder.dev/docs/integrations/aws-s3
Anthropic Claude Opus 4.6 is now available in Falcon Builder by NeoSky AI.
Opus 4.6 is built for long-horizon reasoning across large codebases, complex analysis, and multi-step agent workflows. Falcon Builder enables these capabilities through native orchestration, letting teams design, run, and manage long-running agents without losing signal.
With support for a 1M token context window, context compaction for sustained workflows, and 128K output tokens for single-shot reports and deep analysis, Opus 4.6 in Falcon Builder is designed for production-grade AI systems—not demos.
Falcon Builder adds the operational layer: workflow control, observability, and secure execution—so frontier models can be deployed with confidence.
Explore Claude Opus 4.6 in Falcon Builder → https://falconbuilder.dev
-
1
Congrats on the launch, looks solid. How are you currently thinking about acquiring early users and gathering feedback?
-
1
Thanks. Sorry for the delay - I just saw your reply. My main focus has been on content marketing. I'm about to start really focusing on direct outreach, though.
-
1
That makes sense, content + outreach is a solid combo.
One thing I’ve seen work well alongside that is treating Reddit as a form of “warm outreach.” Instead of reaching out cold, you’re joining conversations where people are already talking about the problem you solve.
It usually leads to better engagement because the intent is already there.
If you want, I can share a simple way to approach it so it actually brings in users and not just traffic.
-
-
Most AI tools are built for demos.
We built Falcon Builder for production.
Over the past year I’ve been deploying AI agents into real business workflows — sales, healthcare intake, voice routing, reporting automation.
The consistent problem:
It’s easy to build an AI demo.
It’s hard to operate an AI system reliably.
So we’re building Falcon as infrastructure:
• structured tool invocation
• webhook orchestration
• system-level logic
• voice + CRM integrations
• production deployment patterns
Would love feedback from other founders building agent-based systems.
Comment