EpicStaff/EpicStaff

EpicStaff – Self‑hosted visual platform for building, running and auditing AI agent workflows

What it is – EpicStaff is a source‑available (open‑source‑ish) platform that lets you design, execute and monitor multi‑agent AI pipelines. A Django backend powers a node‑based visual editor where each node can run custom Python code, call LLMs/embedders, perform Retrieval‑Augmented Generation, or invoke arbitrary tools. The whole stack runs in Docker (or Podman) and can be deployed on‑premises, giving operations teams full ownership and auditability of the agents they run.

---

Core ideas

Audience	Problem addressed	How EpicStaff solves it
Engineers	Agent flows live in code, hard to hand‑off, no runtime governance.	Provides a Django‑backed execution engine, a persistent session ledger, sandboxed per‑agent permissions, and a Python‑first node model that can be extended with any custom tool.
Operations / Business users	Need to modify or audit flows without writing code; lack visibility into decisions made by agents.	Drag‑and‑drop visual workflow builder, real‑time execution logs, role‑based access control, and the ability to pause for human review.
Enterprises	Security, compliance, and data‑sovereignty requirements.	Self‑hosted deployment, JWT‑based auth, configurable secret keys, and storage of agent memory in PostgreSQL (with pgvector) and Redis.

---

Key technical features (for AI engineers)

• Node‑based UI – Build execution graphs visually; each node can hold Python, an LLM call, a RAG query, or a custom tool.
• LLM & embedder abstraction – Switch between OpenAI, Anthropic, or any self‑hosted model with a unified provider interface.
• Built‑in RAG – Upload documents, create vector collections (naïve or GraphRAG), and query them from any node.
• Persistent agent memory – Short‑term window + long‑term state stored in PostgreSQL + Redis, enabling context across sessions.
• Custom tool integration – Write arbitrary Python functions; the orchestration engine automatically makes them callable from the flow.
• Webhooks & API triggers – Start flows via HTTP POST, allowing external systems (CRM, email, monitoring) to invoke agents.
• Debug & audit – Structured flow messages and execution logs per node, viewable in the UI for traceability.
• Enterprise security – Role‑based permissions, JWT authentication, and isolated Docker volumes for data.

---

Business‑focused value

• Visual workflow builder – Non‑technical staff can assemble automations without touching code.
• Multi‑agent crews – Coordinate several agents to automate whole departments.
• Voice & multilingual support – Interact with agents by speech in many languages.
• Human‑in‑the‑loop – Pause flows for manual approval before critical actions.
• Self‑hosted & auditable – Keeps data on‑prem, provides a full decision trail for compliance.

---

Quick‑start (Docker‑first) — run in minutes

1. Install Git and Docker Desktop (or Podman).
2. Clone the repo and launch (example for macOS/Linux; Windows uses PowerShell):

   git clone -b main https://github.com/EpicStaff/EpicStaff.git && \
   cd EpicStaff && \
   cp src/.env.example src/.env && \
   savefiles="$HOME/savefiles" && \
   sed -i "s|CREW_SAVEFILES_PATH=/c/savefiles|CREW_SAVEFILES_PATH=$savefiles|" src/.env && \
   sed -i "s|^SECRET_KEY=.*|SECRET_KEY=$(openssl rand -base64 48 | tr -d '=+/')|" src/.env && \
   sed -i "s|^JWT_SECRET=.*|JWT_SECRET=$(openssl rand -base64 48 | tr -d '=+/')|" src/.env && \
   docker volume create sandbox_venvs crew_pgdata graph_data crew_config media_data && \
   docker network create mcp-network && \
   cd src && docker-compose up --build
   

3. After the containers are healthy, open http://localhost in a browser to access the visual editor.

Alternative setups – partly‑local services or Podman are documented in the repo’s partly-local-setup.md and podman-setup.md.

---

When you might use EpicStaff

• Automating ticket triage, incident response, or IT service‑desk workflows where ops need to tweak routing rules.
• Building a “digital crew” that gathers data, writes reports, and sends notifications across multiple internal APIs.
• Deploying a secure, auditable RAG‑backed chatbot that must retain conversation history for compliance.
• Any enterprise scenario where AI agents must be owned, edited, and reviewed by non‑engineers while still leveraging custom Python integrations.

---

Getting involved

• Star the repo if you find it useful.
• Contribute via pull requests; see CONTRIBUTING.md for guidelines.
• The project builds on the f‑flow library (by Foblex) for the interactive engine.

---

Links

• 📄 Website
• 📚 Wiki & docs
• 🐞 Issue tracker

---

EpicStaff hides the complexity of multi‑agent orchestration while keeping every decision visible and editable.