Every working build shared to the PromptFlow™ System, whether made with Claude, any other model, or a whole pipeline, becomes a reusable process: searchable, adaptable, and runnable from your command line. Create, reuse, and adapt what already works, instead of regenerating it from scratch.
Most AI tools are one-shot generators: you prompt, you get an output, the work evaporates. The PromptFlow™ System persists working builds as processes in a shared graph, so the next builder starts from something proven instead of a blank prompt.
Nothing enters the graph unless its author shared it. Every process carries its run history, including completion rate, cost, and reuse count, so what you find is pre-filtered by the people who built and ran it.
Search the graph, pull a process, run it locally, and pipe its output into your own tools, all from your terminal. No browser required. The working outputs of other builders' sessions become your starting point.
Close but not exact? Fork a process and refine it in plain language. The system rewrites the steps while keeping what worked. Adapting a proven build is faster and more reliable than regenerating from scratch.
Publish a process and set your price. When others run it, fork it, or compose it into something bigger, royalties flow back to you automatically, at every layer of reuse.
The graph is agnostic about how a build was made. MCP servers, n8n workflows, agents, hosted models, `.promptflow` files, and typed HTTP services all become shareable processes with the same interface, the same run history, and the same royalty treatment for their authors.
Tools surfaced via tools/list, registered as capability nodes. Every operator gains discovery and revenue.
Workflows with declared I/O schemas become composable capabilities. Creators gain distribution and monetization.
LangChain, AutoGen, CrewAI, or custom: any agent with an addressable endpoint. Framework choice is invisible to the graph.
Anthropic, OpenAI, Google, open source: every model API is a provider type. Calls pass through at cost; usage is metered per run.
Authored files that compose any of the above. The universal contributor path. Anyone can build one without infrastructure.
Generic endpoints with request/response schemas. If it's accessible and typed, it composes.
The PromptFlow™ Interpreter runs entirely from your command line. Search the shared graph, run any published process locally, adapt it to your inputs, and publish your own. Same auth, same royalty flow, no browser needed.
$ pip install promptflow-interpreter
$ promptflow login
$ promptflow search "shopify listing"
1. shopify-listing ★4.8 $0.12/run
2. listing-translator ★4.6 $0.08/run
$ promptflow run shopify-listing -i img.jpg
$ promptflow publish my-processThe process layer isn't a rebrand. The patent filed with a 2023 priority date describes exactly this: an interpreter, a shared database of processes, and reuse of what already works. Both the patent and the PromptFlow™ trademark have received Notices of Allowance from the USPTO. Builders keep ownership of what they publish and earn royalties when it is reused.
The system is patent-allowed and the mark is protected. Builders keep ownership of what they publish and earn when it is reused. The graph is built by the people who own it.