For teams taking AI into production
Design AI workflows visually, run them on your own infrastructure, and trace every step. A model is called only where a step needs one, so runs stay predictable and costs stay down.
Why workflows
Agents re-plan on every run. A workflow decides the steps once and reuses them, so every run is predictable.
Every run follows the same steps, so you can compare results side by side, test a change before it goes live, and see exactly what got better.
The workflow captures how your team does the work. It’s versioned and reviewed like code, so the know-how stays with your company, not in chat logs.
Each AI step does one narrow job, so it can use a smaller, cheaper model with a shorter prompt. You pay for a large model only where it really matters.
Use a free-form agent for open-ended steps and fixed steps everywhere else. You choose the balance, step by step.
When a customer, auditor or manager asks why the system did something, you can point to the exact step and data behind it.
| Aspect | Free-form agent | Faraflow workflow |
|---|---|---|
| When a run fails | Run the whole thing again. | Retry from the step that failed. |
| Changing behavior | Edit the prompt and hope. | Change one step and test it before it goes live. |
| Choosing models | One model for the whole job. | Any of 100+ providers, chosen per step. |
| Using it elsewhere | Lives in a chat or script. | Publish it as an API. |
Run the whole thing again.
Retry from the step that failed.
Edit the prompt and hope.
Change one step and test it before it goes live.
One model for the whole job.
Any of 100+ providers, chosen per step.
Lives in a chat or script.
Publish it as an API.
Capabilities · Live today
Everything here works in Faraflow now. Anything still to come is on the roadmap, marked that way.
DriveUp is an AI visibility analysis tool for self-storage operators integrated with multiple LLM providers and many other sources. Its report workflows run on the Faraflow engine, in production.
selfstorageapps.com
Every node shows its latest outputs right on the canvas.
AI, code, logic and integration nodes, plus your own custom nodes.
See each run node by node, and load any past run back onto the canvas.
Step through a node’s outputs, config and connections, upstream and downstream.
Keep real inputs as test cases and rerun the workflow against them.
Open source, so you can see and run everything.
Backed by Alokin, so you’re never on your own.
The source is open. Inspect what your workflows run on, and change it if you need to.
On your own infrastructure. You decide where it runs and where your data lives.
Open code means you’re not dependent on a single vendor to keep it running.
Alokin supports Faraflow deployments for teams that need someone to call.
Alokin’s engineers can design and build workflows for your business and connect them to your systems.
Alokin maintains and extends Faraflow; the roadmap below shows what’s next.
Developed and commercially supported by Alokin, a software consultancy delivering software since 2003, including SaaS, web/mobile development, APIs, AI/NLP, e-governance and enterprise systems.
Faraflow grew out of the open-source PySpur project and has been heavily extended since.
Roadmap
We’re constantly building and improving Faraflow. Here’s what’s coming next.
Your use case can reshape this list.
Tell us what matters to you.
FAQ
What people ask us most about Faraflow. Anything else, ask us directly below.
Faraflow is an open-source platform for building AI agents and workflows, developed and supported by Alokin. You design a workflow visually in the studio, run it with the engine on your own infrastructure, and trace every step of every run.
Yes. The source code is open, so you can read it, run it yourself and change it. Faraflow grew out of the open-source PySpur project and has been heavily extended since.
Yes. Faraflow runs on your own infrastructure, so you decide where your workflows run and where your data lives.
An agent plans its work again on every run. In Faraflow the steps are decided once, a model is called only where a step needs one, and every run follows the same path. You can still use a free-form agent inside a step where the task is open-ended.
LangChain and LangGraph are code libraries: they give developers building blocks for writing agents and graphs in Python or JavaScript. Faraflow is a complete studio and engine. You design the workflow visually, run it on your own infrastructure, and get run traces, test cases, retries and publishing as an API built in, so the whole team can see and change how a workflow works. Code nodes let developers add custom logic where they need it.
Faraflow connects to 100+ model providers, and you choose the model per step: a smaller, cheaper model for routine steps and a larger one where judgment matters.
Yes. DriveUp, an AI visibility analysis tool for self-storage operators, runs its report workflows on the Faraflow engine in production.
Alokin, a software consultancy delivering software since 2003. Alokin maintains Faraflow, offers commercial support, and can design and build workflows for your business.
Tell us what you are trying to solve, and we will show you how it could work as a workflow.
Meet us in person
Come by for a live walkthrough of the studio and the engine, or bring the workflow you have in mind.
Prefer email? rajeev@alokin.in