# AI use cases for Exo (exo-explore/exo) Repo: [exo-explore/exo](https://github.com/exo-explore/exo) | Use cases: 3 | Page: https://dedupedai.com/use-cases/exo-explore-exo/ exo connects your devices into one AI cluster so you can run models larger than any single device could hold, and run them faster as you add devices. Devices on the same network discover each other automatically, exo splits the model with topology-aware auto parallel and tensor parallelism on an MLX backend, and supports RDMA over Thunderbolt 5. Each device exposes a dashboard and an API compatible with OpenAI Chat Completions, Claude Messages, OpenAI Responses and Ollama. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Pool several Macs to run a model none could run alone: https://dedupedai.com/use-cases/exo-explore-exo/pool-several-macs-into-one-ai-cluster/ - Use your existing AI tools with a local Exo cluster: https://dedupedai.com/use-cases/exo-explore-exo/point-existing-ai-tools-at-an-exo-cluster/ - Run a test cluster next to your main one: https://dedupedai.com/use-cases/exo-explore-exo/run-a-separate-test-cluster-with-exo/ ## How to pool several Macs to run a model none could run alone with Exo Repo: [exo-explore/exo](https://github.com/exo-explore/exo) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Install Exo on each Mac and let them discover each other and split a large model across their memory. **Who it is for:** Someone with two or more Apple silicon Macs who wants to run a model bigger than one machine's RAM. **The problem:** Large open models need more memory than a single desktop or laptop has. **The outcome:** One cluster that runs the model across all Macs, managed from the dashboard at localhost:52415. ### Steps 1. Install the macOS app from EXO-latest.dmg or with `brew install --cask exo` 2. Open EXO and allow Local Network access so it can find your other devices 3. Install EXO on the other Mac, put both on the same Wi-Fi or Ethernet network or connect them with a Thunderbolt cable, and make sure both run the same version 4. Watch the new Mac appear in the dashboard within a few seconds 5. Pick a model in the dashboard and chat; Exo's topology-aware auto parallel decides how to split it ### Prompt to paste into your agent ```text Help me set up exo-explore/exo on my Macs: install with `brew install --cask exo`, check they run the same version, and troubleshoot if one does not show up in the dashboard. ``` ### Caveat Devices running different Exo versions do not join each other, so update every Mac together. Risk, system: The optional Thunderbolt network setup installs a system service that switches the Mac to an Exo network location; custom IP, DNS or proxy settings do not apply while it is in use. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/exo-explore-exo/pool-several-macs-into-one-ai-cluster/ ## How to use your existing AI tools with a local Exo cluster with Exo Repo: [exo-explore/exo](https://github.com/exo-explore/exo) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Point clients that speak the OpenAI, Claude Messages or Ollama APIs at Exo instead of a cloud provider. **Who it is for:** A developer who already uses OpenAI or Ollama clients and wants to run them against local hardware. **The problem:** Switching to local models usually means rewriting integrations for a new API. **The outcome:** Your existing tools talking to models running on your own cluster through the API on port 52415. ### Steps 1. Start Exo on each device (app or source) so the cluster is up 2. Note the API and dashboard address each device provides, http://localhost:52415 3. Configure your client's base URL to that address using its OpenAI Chat Completions, Claude Messages, OpenAI Responses or Ollama mode 4. Load a custom model from the Hugging Face hub if the one you need is not listed 5. Watch the dashboard while requests run to see which devices serve them ### Prompt to paste into your agent ```text Using exo-explore/exo, configure to send its requests to my exo cluster at http://localhost:52415 through its OpenAI-compatible setting, then run a test prompt and confirm the response came from the cluster. ``` ### Caveat This card is written from the exo-explore/exo README, not from a test run by Deduped. Before you start, make sure you have For the macOS app: a Mac with Apple silicon and macOS Sequoia 15.1 or later. Check the repo for newer instructions before you rely on these steps. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/exo-explore-exo/point-existing-ai-tools-at-an-exo-cluster/ ## How to run a test cluster next to your main one with Exo Repo: [exo-explore/exo](https://github.com/exo-explore/exo) | Difficulty: Intermediate | Time: 20 minutes | Basis: From the README Give devices their own cluster name so a test setup does not join your main Exo cluster. **Who it is for:** A home lab user trying a new Exo build or model without disturbing the cluster everyone uses. **The problem:** By default every device with the same Exo version joins the same cluster. **The outcome:** Two independent clusters on one network, each with its own name. ### Steps 1. Check each device's version with `uv run exo --version` when running from source 2. Start test devices from source with `uv run exo --namespace my-cluster`, or set EXO_ZENOH_NAMESPACE 3. In the app, set Settings, Cluster, Cluster name instead 4. Keep devices with the same custom name on the same Exo version 5. Check ~/.exo/exo_log/exo.log for devices that were seen but could not join ### Prompt to paste into your agent ```text With exo-explore/exo running from source, start a separate test cluster with `uv run exo --namespace my-cluster` on , confirm in the log which namespace each one uses, and make sure my main cluster is untouched. ``` ### Caveat An app install and a source checkout number versions differently, so they only join each other with the same custom cluster name. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/exo-explore-exo/run-a-separate-test-cluster-with-exo/