9
AI PROVIDERS
PERSONAL PROJECT
Your compass.
Your agents.
One team.
A full-stack environment where I learn how agents function and work safely. Built out of curiosity, not for a market.
AI providers: ANTHROPIC, OPENAI, GOOGLE, MISTRAL, GROQ, DEEPSEEK, PERPLEXITY, NVIDIA, MOONSHOT
WHY I BUILD THIS
Not a startup.
My personal lab.
Project Fusion is my playground and my lab in the shape of a full-stack environment. I started out to explore what Claude Code can do. Where I ended up is a productivity app with its own harness, sandboxing and trust core.
That deepened how I work with AI. The harness I built myself shows me what matters in tooling and security. That is the purpose of this project, and it is already fulfilled.
STATUSOngoing, at my own pace, no deadline and no release date.
PURPOSEA learning environment and a tool for myself. Not a product.
IF RELEASEDThen as open source. I do not run a service and I am not going freelance.
MY FOCUSMy job as Solution Architect for Automations & AI at SKOUZ. Project Fusion stays my lab alongside it.
CORE MODULES
Many modules,
one system
A workspace built for humans and their agents. Every module is a tool both can use, in one system and on the same data.
01
Routines · Workflows
Create multi-step flows with connected nodes. Your agent is just one block among many. Chain tools and modules into repeatable workflows.
02
Impulse · Automation
Scheduled triggers that run your routines automatically. In the background, no human in the loop. Define once and benefit repeatedly.
03
Tracker
The task board for you and your agent. Tasks, status and progress in one place. Managed by you or filled automatically by agents and routines.
04
Mail & Planner
Full email client and calendar for Microsoft and Google, on board. Agents may work with the modules if you allow it. You keep the overview.
05
Scout · AI Evaluator
Rate and compare products, jobs, companies, hotels and more. By your criteria and your weights. Have your gut feeling checked and backed up.
06
Multi-Model Chat
Your chat interface is your bridge. Switch agents, pick the right model for the task and control permissions at any point in the session.
APP PREVIEW
Built for
humans and agents.
One workspace for both sides. You work here, your agents work alongside you. Every view is built for that.
THE AGENT HARNESS
Security learned
by building it
The harness is the environment an agent thinks and acts in. Building one yourself shows quickly which limits an agent needs and where tooling gets dangerous. These blocks are the result.
01Isolated SandboxEvery agent runs in its own contained environment. Nothing leaks out.
02Permission GatesYou grant and revoke capabilities per module. Nothing happens without consent.
03File AccessScoped read/write access to exactly the files an agent needs.
04TerminalDirect shell access for real work on files and commands.
05NetworkingControlled outbound access: APIs, web, services.
06Event SystemEvery run, every tool, every decision emits an event. Monitoring is the natural consequence.
Got
curious?
Project Fusion keeps growing, step by step and without a deadline. What takes shape here shows how humans and agents can work safely in one system: with clear boundaries, traceable decisions and a human at the helm.
Questions about the project, experiences with AI agents to share, or just want to talk shop? I look forward to the exchange.
OPEN FOR QUESTIONS · EXCHANGE · FEEDBACK