I kept building AI agents and hitting the same wall: the agent worked, but getting it in front of users meant building auth, multi-tenancy, a chat UI, conversation storage, access control. Months of infrastructure before anyone could use the thing.
AgentFlow is my attempt to solve this. Connect any HTTP/SSE endpoint and get a production-ready interface with complete tenant isolation. Your LangChain agent, n8n workflow, custom Lambda, whatever speaks HTTP — plug it in, you're live.
The real use case: you're an AI consultant or agency with multiple clients. Client A needs a LangChain contract analyzer. Client B needs an n8n refund bot. Without something like this, you're managing two deployments, two auth systems, two codebases. With AgentFlow, it's one portal. Route each client to their own backend. Complete data isolation. Same professional frontend.
Or you're a SaaS company adding AI features at different tiers. Free users get GPT-3.5, Pro gets GPT-4, Enterprise gets your fine-tuned model. Create groups, assign endpoints, done.
Stack: Next.js, Supabase, Row-Level Security. Config is YAML-based with a UI wizard option.
This started as a tool I needed for my own consulting work. AGPL-3.0, self-hostable, free forever. Happy to answer questions.
I kept building AI agents and hitting the same wall: the agent worked, but getting it in front of users meant building auth, multi-tenancy, a chat UI, conversation storage, access control. Months of infrastructure before anyone could use the thing.
AgentFlow is my attempt to solve this. Connect any HTTP/SSE endpoint and get a production-ready interface with complete tenant isolation. Your LangChain agent, n8n workflow, custom Lambda, whatever speaks HTTP — plug it in, you're live.
The real use case: you're an AI consultant or agency with multiple clients. Client A needs a LangChain contract analyzer. Client B needs an n8n refund bot. Without something like this, you're managing two deployments, two auth systems, two codebases. With AgentFlow, it's one portal. Route each client to their own backend. Complete data isolation. Same professional frontend.
Or you're a SaaS company adding AI features at different tiers. Free users get GPT-3.5, Pro gets GPT-4, Enterprise gets your fine-tuned model. Create groups, assign endpoints, done. Stack: Next.js, Supabase, Row-Level Security. Config is YAML-based with a UI wizard option.
This started as a tool I needed for my own consulting work. AGPL-3.0, self-hostable, free forever. Happy to answer questions.
GitHub: https://github.com/connorbell133/Agentflow