Introduction
When I first started using AI agents, I felt like I was becoming ultra-productive. But pretty quickly, I realized the work had not disappeared. It had moved.
Instead of doing every task myself, I was spending more time supervising agents, checking their work, recovering from mistakes, and reconstructing context across tools and sessions. I could not fully delegate complex work because I could not reliably preserve intent, decisions, plans, progress, and outcomes over time.
As my workload scaled with the new capability, the bottleneck became coordination. I needed a way to manage long-running AI-assisted work, track what happened and why, and maintain a consistent operating system across different models, providers, and projects.
So I built Festival.
Festival is a local-first workflow system for organizing complex work with AI agents. It helps preserve context, structure execution, and make AI-assisted work easier to resume, review, explain, and improve. It allows me to keep a consistent workflow overtime across provider harnesses, while retaining full control over my data, context and work output.
The Problem
AI agents are powerful, but they are difficult to trust with complex work over time.
Most tools focus on generating outputs, but they do not give you a reliable way to manage the surrounding work: the plans, decisions, review points, context, artifacts, and next actions that make the output useful. Agent harness providers help somewhat with this, but their data formats are not consistent, their pricing/features are constantly changing and their solutions are rarely portable between harness providers.
As a result, people using AI for serious work often end up with scattered notes, unclear project state, repeated context rebuilding, and a lot of manual supervision. The more capable the agents become, the more coordination work the human has to manage.
The main challenge is not just getting AI to produce work. It is directing AI reliably across long-running workflows without losing control, context, or accountability.
The Solution
Festival gives AI-assisted work a durable structure.
It organizes work into clear campaigns, phases, sequences, tasks, goals, reviews, and artifacts so both humans and agents can understand what is happening, what has already happened, and what should happen next.
Instead of treating every AI session like a fresh conversation, Fest helps preserve the operational memory around the work. It lets you continue across tools, models, and sessions without constantly rebuilding context from scratch.
The goal is simple: make complex AI-assisted work easier to delegate, monitor, resume, and explain.
Key Features
- Structured workflows: Festival turns messy, long-running work into organized plans
- Persistent context: Fest keeps track of decisions, progress, artifacts, and next actions so work can continue across sessions without starting over.
- Provider-independent execution: Festival gives you a consistent workflow layer across AI tools and models, so your process does not depend on one provider or interface.
- Easy navigation: Festival provides ergonomic navigation to make jumping around nested context directions and multiple projects seemless with minimal chance of collisions
Conclusion
I makes it possible to do more work than ever, but more capability creates a new management problem. Without structure, AI-assisted work becomes harder to trust, harder to explain, and harder to scale.
Festival is my attempt to solve that problem: a workflow layer for people doing complex, long-running work with AI agents.
If you are building with AI agents and running into the same coordination problem, try Festival, read the docs, or follow the project on GitHub.
It's a very deep project based on applied theory in multiple displines, I'd recommend installing it and trying it rather than fully attempting to understand what it does. Most of the commands are for agents, as a human user you'll mostly use camp create, cgo(camp go), csw(camp switch), fgo (fest go), fls(fest list) and a handful of other commands when you feel the need to use them. Don't be intimidated by the depth or breadth of the system, it's designed to be flexible and customizable to adapt to any workflow.