01
Desktop app
Install it, sign in, and the agent is in its own window. No terminal, no setup after the installer.
Install the app →The local-first agent framework that learns.
An open-source AI agent framework with persistent memory, tool use for web, code, and files, and a living dot that shows what your agent is doing. Install the desktop app or self-host it on your own server; your data stays with you.
Open source · macOS, Windows & Linux
One agent. Every capability wired in. Memory Web search Sandboxed code Browser Files Self-learning Multi-agent
Three ways to run
Local-first by design. Yours to run.
One agent, local-first by design. Run it on your machine, deploy it to your own server, or build on the framework and ship your own agents.
02
Self-host it
Install VAF as a host on your own server. Your data, your keys, your infrastructure, fully under your control.
Read the docs →03
Build on the framework
VAF is a framework and harness in one. Extend it, wire in your own tools over MCP, and ship your own agents.
Explore the framework →No providers to wire up, nothing to host. The Veyllo API gives VAF every model and connection it needs behind one key.
See it work
The main agent
You ask in plain language, and the main agent plans the work and delegates each part to the specialist sub-agents below.
Why Veyllo
An agent that remembers, acts, and improves.
What makes it worth running locally.
Remembers across sessions
Vector-backed memory lets the agent recall people, projects, and preferences across sessions, so context carries over instead of resetting on every restart.
Runs on a schedule
Tell the agent what to do and when, once at noon on the 23rd or every day at 17:00. It runs the task on its own at that time and reports the result back.
Learns your tools
Connect a new or custom tool and the agent learns it by trial and error. It uses the tool, reads the results of its own attempts, and adjusts until it can use the tool on its own.
Stays on your machine
Runs on your machine, in its own desktop window. Your data and keys stay where you put them.
Runs multi-step work on its own
The agent plans a task into steps and runs them on autopilot, coordinating each step across the run. It watches progress and adjusts the plan as things change, so the work stays on track.
Local-first AI agents
What it means to run an agent on your own machine, and why local-first matters for privacy, cost and control.
Read the explainer →Veyllo API · preview
Every model VAF needs, behind one key.
VAF runs on models and the services it plugs into. Normally you sign up with each provider and wire in keys yourself, or host models on hardware you may not have. The Veyllo API replaces all of that with one key.
✓
One key instead of many. Every model and provider VAF talks to, behind a single API. No scattered accounts or keys to manage.
✓
Nothing to host locally. No GPUs, no self-run model servers. Compute and model hosting run on Veyllo.
✓
Built for production. Keys, rate limits and usage metering from day one.
Install
Bring the agent to your desktop.
Clone the repo and run one installer. It sets up the container runtime, Python and Node for you. Pick your platform for the exact steps.
Open development
Follow the build on GitHub.
Read the code, track the roadmap, and file issues. Development happens in the open.
Active
development
Issues
open & tracked
Roadmap
public
Put the agent to work.
Install the desktop app, or build on the Veyllo API.