Repolode
Submit
davemessew

davemessew/AgentFlux

Self-hosted agentic AI platform with a visual workflow editor, autonomous agents and local document storage, run via Docker.

View on GitHubWebsite
Stars
199
Forks
3
Open issues
0
Last commit
9 days ago

What is it?

AgentFlux connects documents, data sources and language models in a visual workflow editor. Users can build each step by hand or describe a task and let an agent generate the workflow. The application runs in a browser interface on your own computer or server, and workflows, documents, credentials and run logs are kept in a local Docker volume.

The platform includes 17 preconfigured workflow steps, eight templates, branching, undo/redo and JSON import/export. Agents can be given a task, a set of allowed sources and a model, then started manually or on a schedule. It can also analyse a screen recording with a vision-capable model and turn the observed actions into an editable workflow.

Models can come from Ollama, LM Studio, vLLM or any OpenAI-compatible endpoint. Files such as PDF, DOCX, XLSX, CSV, JSON and text can be imported, and folders, HTTP endpoints and approved programs can be connected. Cloud model calls and external HTTP POST steps require explicit approval. The project has no built-in login and is designed for a single user.

Key features

  • 01Visual workflow editor with 17 preconfigured steps, templates, branching and undo/redo
  • 02Agents that run tasks manually or on a schedule using selected sources and a model
  • 03Video analysis that turns a screen recording into an editable workflow
  • 04Connects to Ollama, LM Studio, vLLM and OpenAI-compatible model endpoints
  • 05Imports PDF, DOCX, XLSX, CSV, JSON and text files into a local knowledge base
  • 06Runs with Docker Compose; data persists in a named volume

Use cases

  • Running document processing and data cleaning pipelines on your own machine
  • Automating recurring tasks with a local model such as Ollama
  • Turning a recorded screen process into a repeatable workflow
  • Keeping sensitive files and model calls within your own infrastructure

Quick start

  1. 1

    Download the two release files (image tarball and compose.yaml) into the same folder from the v0.2.0 release page

  2. 2

    Load the prebuilt Docker image

    docker load -i agentflux-0.2.0-linux-amd64.tar.gz
  3. 3

    Start AgentFlux with Docker Compose, then open localhost:4310

    docker compose up -d --wait

Things to know

  • Prebuilt Docker image targets Linux/amd64 only.
  • No login; designed for a single user and needs extra access control for shared use.
  • Docker image does not include an AI model; you must connect your own model server or provider.
  • Video-to-workflow analysis is not in the prebuilt v0.2.0 image and requires building from source.

In collections

Open-source AI agent frameworks & appsSelf-hosted appsRun LLMs locally

Topics

#ai-agents#workflow-automation#self-hosted#docker#ollama#local-llm#visual-programming#workflows

FAQ

What is AgentFlux?
Self-hosted agentic AI platform with a visual workflow editor, autonomous agents and local document storage, run via Docker. AgentFlux connects documents, data sources and language models in a visual workflow editor. Users can build each step by hand or describe a task and let an agent generate the workflow. The application runs in a browser interface on your own computer or server, and workflows, documents, credentials and run logs are kept in a local Docker volume.
Is AgentFlux free to use? What license does it have?
AgentFlux is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is AgentFlux still maintained?
Actively maintained. The last commit to AgentFlux was on Oct 30, 2026.
What can I use AgentFlux for?
Running document processing and data cleaning pipelines on your own machine Automating recurring tasks with a local model such as Ollama Turning a recorded screen process into a repeatable workflow Keeping sensitive files and model calls within your own infrastructure
How do I install AgentFlux?
Download the two release files (image tarball and compose.yaml) into the same folder from the v0.2.0 release page Load the prebuilt Docker image: docker load -i agentflux-0.2.0-linux-amd64.tar.gz Start AgentFlux with Docker Compose, then open localhost:4310: docker compose up -d --wait
Where can I get AgentFlux?
The source code and installation instructions are on GitHub: https://github.com/davemessew/AgentFlux

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects