The Distributed Open Llama Network
Turn the computers you already own into a personal AI cluster.
Your most capable computer should not become useless the moment you leave your desk. Dollama connects the hardware you already own so it can work together.
Keep models and agents running on an always-on home server. Send work to whichever of your machines is best equipped for it.
Run tasks across your machines in parallel. Continue working when one machine is busy or unavailable.
Your devices remain independent computers. Dollama gives them a shared coordination layer.
Dollama is more than a remote connection to one model. It coordinates different models and computers so each part of a task runs where it fits best.
A powerful planner model may be used briefly to understand a difficult request, break it into manageable work, and choose the right tools. Smaller worker models can then perform most of the task efficiently.
When several investigations can happen independently — and you have more than one machine to run them on — they can run at the same time. Helper and sub-agent work can distribute across your cluster.
If a worker stops early, Dollama can prompt it to continue. If an attempt fails, the system can retry or route the work elsewhere. Individual models can still make mistakes. The whole system does not have to fail with them.
Smaller models do most of the work. Stronger models help them succeed. A few powerful planner nodes can improve the success of a much larger pool of ordinary machines.
Dollama starts with the resources available to you.
Your requests prefer your own hardware first. A laptop on battery may use a desktop at home.
Next, machines shared by groups you belong to. A smaller local model may receive a plan from a stronger group node.
Independent work may be distributed across several workers. The goal is to complete the task well using the resources available.
You do not need an expensive GPU to benefit from Dollama. Even a single modest machine — or one with no GPU at all — can use the open network.
Stronger planner models. Additional worker and helper capacity. Models your hardware can't run. Concurrent execution across other people's machines.
Your computer does not need to run every model itself. People with a whole suite of machines get more from their own hardware; people with one machine still get coordinated models, recovery, and free network capacity.
Anyone can submit work to the open network without payment.
No subscription. No per-token fee. No network usage fee. No paid membership fee. Contribution affects priority rather than permission.
People who contribute useful compute receive faster access when the network is busy. Requests from people who contribute less may spend longer in the queue during periods of sustained demand.
Compute itself is not free. Nodes consume real electricity, hardware, bandwidth, and time. Those resources are supplied by people and organizations that choose to participate. Dollama coordinates that shared capacity without selling it back to the community one token at a time.
Dollama is infrastructure, not another closed AI application.
Point a compatible coding tool, agent, chat client, or application at a local Dollama endpoint. Dollama handles where inference runs while tools, credentials, and project access remain on your machine.
Coding assistants. Agent frameworks. Chat applications. Research and RAG systems. Automation tools. Private organizational software. Products built by other teams. You should not have to replace your workflow just to change where the intelligence runs.
Dollama is open source. You can inspect it, run it, build on it, adapt it, and contribute improvements.
Node discovery and routing. Planner, worker, and helper roles. Context handling. Retries and recovery. Contribution accounting. Compatible API endpoints. Relays. Reference applications.
A clearly separated open protocol that other teams can implement independently. Clients, nodes, schedulers, runtimes, relays, private networks, applications. The Dollama implementation should be an easy way to join the network — not the only way.
No suspiciously sparkly privacy claims. Just an honest description of where trust still exists.
Work runs only on nodes you own. If none of your own machines are available, the request fails rather than silently falling through to a group or the public network.
dollama privateWork may run on your own nodes plus nodes shared by groups you belong to. Like Private, Group fails closed: if no eligible own-or-group node is available, the request fails rather than falling through to the public network.
dollama groupWork may run on eligible community contributors. Use this mode only for tasks you are comfortable sending to an untrusted computer.
dollama networkA machine that executes a prompt may be able to inspect that prompt. Tunnel traffic still passes through relay infrastructure. Dollama is not funded by collecting, profiling, or monetizing your prompts. That does not remove the need to understand which machines execute them.
curl -fsSL https://dollama.net/install.sh | shRequires Ollama installed with qwen3.5:9b pulled if you plan to contribute.
irm https://dollama.net/install.ps1 | iexOpen Windows PowerShell (not Command Prompt) and paste the line above — irm and iex are PowerShell commands and won't work in cmd.exe. The installer downloads dollama, sets up Ollama, adds a Start Menu shortcut, and opens the dashboard for you.
Aliases disabled? Use Invoke-RestMethod https://dollama.net/install.ps1 | Invoke-Expression instead.
Requires Ollama installed with qwen3.5:9b pulled if you plan to contribute.
Windows may show a SmartScreen warning — click "More info" then "Run anyway". This is normal for unsigned open-source software. Troubleshooting
The installer already launched dollama in your system tray. Choose your mode:
dollama private # your own machines onlydollama group # your own + your groups' machinesdollama network # the full open network (and contribute)dollama launch claudeThat's it. You're running on community compute. Dollama exposes an Anthropic Messages API endpoint at localhost:11435. Any tool that supports a custom base URL can use it — Claude Code, Continue, Aider, or your own scripts.
~/.dollama/bin (macOS/Linux) or %LOCALAPPDATA%\dollama (Windows). On macOS it also creates ~/Applications/dollama.app for Spotlight/Launchpad; on Windows it adds a Start Menu shortcut. All platforms open the dashboard after install.
Read the script source.
After extracting on Windows: move dollama.exe to %LOCALAPPDATA%\dollama, add that folder to PATH, then run dollama install-app and launch via the Start Menu → Dollama shortcut (not by double-clicking the zip download).
curl -fsSL https://dollama.net/dl/latest/checksums.txt
sha256sum -c checksums.txtgit clone https://github.com/notangrywaffle/dollama.net.git
cd dollama.net/cli && make buildFull source code on GitHub. Relay, CLI, installer, and this website — all public.
Models change. Providers change. Prices change. Machines fail. Companies disappear. Your access should not disappear with them.
Models should remain replaceable. Applications should remain independent. Relays should remain replaceable. The network should exist for as long as people continue to run it.