Self-Hosting AI Relay Setup
Last updated 2026-09-19What's new
- Buzz (a free, open-source app) now lets you create and edit videos using AI agents like GPT-6 Astra (a video editing AI) without needing traditional editing software.
- ElevenLabs MCP (a creative toolkit) allows AI agents to generate and control voice, music, images, and videos, making it easier to create content.
- You can use AI agents like Zeus (a task manager AI) to create YouTube Shorts and thumbnails, then have another agent like Astra (a quality assurance AI) check the final product.
- Buzz is self-hosted, meaning your data stays private and you can back it up or move it anywhere you like.
- A new tool called Bass (a Rust program) creates a knowledge graph (a map of your information) for your rules, behaviors, and projects, making AI assistant Claude (a chatbot) remember everything between sessions, unlike the static claude.md file.
- Bass uses hooks (automatic code snippets) and an MCP server (a bridge that pulls data on demand) to inject relevant rules and context into conversations, saving on token usage (the data Claude processes).
- Bass offers two tiers: global (follows you everywhere) and workspace (specific to a project), with a dashboard to manage rules and a handoff system to track tasks and progress.
- Unsloth (a free, open-source tool) makes it easy to run and train AI models locally, with a simple interface and support for various hardware and remote access.
- Diagram Design (a plugin for AI agents) helps create clear, professional diagrams like flowcharts and architecture charts, and can be used with agents like Hermes (a cloud-based AI assistant).
- Obsidian Skills (a tool for AI agents) allows agents to interact with Obsidian (a note-taking app using markdown), turning it into a knowledge base or wiki for the agent.
- Buzz (an open-source Slack alternative) is designed for AI agents and humans to work together, with a focus on privacy, security, and self-hosting.
- GrokBot (a new AI tool for non-tech users) can work continuously, even when your computer is off, and perform tasks like managing your LinkedIn inbox, but it may trigger security alerts.
- GrokBot can create and manage multiple AI bots for different tasks, like editing videos or creating content, but the quality may vary.
- Buzz (another AI tool) offers more flexibility and control, allowing users to choose their AI model, access open-source code, and own their data, but it requires server setup.
- Both GrokBot and Buzz have pros and cons, with GrokBot being easier to use but more restrictive, while Buzz offers more customization but requires technical setup.
- Zerp is a new, free app by Spotify that lets you use up to 50 coding agents (AI tools that help write code) at once, all working on the same project without getting in each other's way.
- You can switch between different coding agents, like Claude Code, Codex, or Gemini (all AI tools that help with coding), even in the middle of a task, keeping all your work intact.
- Zerp runs these coding agents in the background using T-mux (a tool that lets you run multiple terminal sessions at once), so they keep working even if you close the app.
- It also shows you how much you're using each AI tool, has a built-in web browser to test your work, and lets you manage all your coding agents in one place.
- Buzz (a chat app with AI features) now lets you use AI models for free by sharing spare computer power with a trusted community, no monthly fees or API keys (a code that lets software talk to each other) needed.
- You can use AI models others in your community are running, combine answers from multiple models, or even split a huge model across several computers.
- Your questions and answers stay private, going directly between computers in your community, but the person whose computer answers can see what you sent.
- To start sharing, enable "share compute" in Buzz's settings, download a model, and create an agent (a tool that uses the shared AI power) pointing to that model.
Key points
What it is
- **Self-hosting an AI relay** means running your own AI agents (programs that act on your behalf) on servers you control, instead of using cloud services like AWS (a cloud computing platform).
- It gives you full control over your data, security, and costs, acting as a middle layer that routes requests between your AI and the tools it uses, like email or calendars.
- This setup is crucial for businesses that rely on AI, as it ensures your AI agents are always online and reachable, and it lets you manage security and privacy risks directly.
- It also provides autonomy, allowing you to route tasks smartly and control how your AI runs, where your data lives, and how much you pay.
How to use it
- **To self-host a Buzz AI relay (server)**, start by signing up for the Buzz desktop app (a window into a server) and creating an account.
- Copy your npub (your Buzz identity) and paste it into a deploy button on a hosting service like Railway (a hosting platform), which handles the complex setup and gives you a URL for your relay.
- You can add a custom domain by configuring DNS records with Cloudflare (a web infrastructure and website security company) or your domain host.
- Enable “share compute” to let other members run agents on your machine, or push code to remote repositories hosted on your relay, integrated with GitHub (a platform for hosting and collaborating on code).
Watch out for
- **Treating a self-hosted relay like an open playground** can be a big mistake, as anyone you let in can upload files, generate content, and use your compute resources.
- **New members can join your relay unless you lock it down**, so it's important to restrict access immediately, especially for business relays.
- **Always check your resource usage**, as you're paying for what members consume, and costs can add up quickly.
- **Consider using Block's hosted relay for open communities** and self-hosting for anything private or business-related, to balance convenience and control.
Tools named
- Buzz (a messaging and file-sharing platform), Railway (a hosting platform), Cloudflare (a web infrastructure and website security company), GitHub (a platform for hosting and collaborating on code)
Lesson 1: What is Self-Hosting AI Relay Setup and why it matters
Self-hosting an AI relay setup means running your own AI agents (programs that act on your behalf) on servers you control, rather than relying only on cloud services like AWS. The term "relay" refers to the middle layer that routes requests between your AI and the tools it uses, like email or calendars. This setup matters because agents need to stay online and reachable 24/7 to be useful—a private agent running on your own hosting gives you that always-on capability from your phone or anywhere else.
Why does this matter for AI development? First, it's an identity choice, not just a tech one. If AI is core to your business, infrastructure is not optional. You can run agents locally for full control, but then you're responsible for keeping them up; hosting on a VPS (virtual private server) or dedicated platform makes it cheaper and more reliable. Second, self-hosting lets you manage security and privacy risks directly—you decide what access your agent gets to sensitive systems, which is critical for power users who understand the threats.
The practical benefit is autonomy. Instead of relying on a vendor's API (interface for connecting programs) and burning cash on per-request fees, you can route tasks smartly. For example, you might give simple, retryable jobs to cheaper models and reserve expensive judgment for complex ones. Ultimately, self-hosting gives you control over how your AI runs, where your data lives, and how much you pay—making it a foundational choice for serious AI builders.
Sources
- 2026-06-29 — Your Own AI Agent in 3 Minutes OpenClaw Hostinger Setup
- 2026-06-20 — Anthropic Says That AI is Improving Itself
- 2026-01-03 — The AI Choice You’ll Regret in 2026
- 2026-07-22 — Stop Burning Cash on Gemini API The New Routing Strategy!
- 2026-02-27 — The NEW Nano Banana 2 + Antigravity Destroys Every AI Image Tool
- 2026-01-27 — Set Up Clawdbot on a VPS in Minutes (no mac mini)
- 2026-07-09 — Matrix Multiplication Wars CPU vs GPU vs TPU vs LPU
- 2026-05-18 — 9 biggest startup ideas right now (AI, B2C, mobile etc)
- 2026-07-12 — The Agentic Web and the Bazaar Era of AI - Ramesh Raskar, MIT Media Lab
- 2026-07-31 — How to Build AI Agents that Actually Work (NO CODE)
- 2026-05-13 — Claude Makes Dashboards Too Easy. Thats the Problem.
- 2026-05-12 — The 1M+ Solo AI Agent Business (Full Course)
- 2026-07-19 — Boris Cherny - The Scaling Trap More Companies Will Hit!
- 2026-06-21 — This Is The First Real Shape Of AGI Fusion Agents
Lesson 2: How to use Self-Hosting AI Relay Setup: step-by-step
Self-hosting a Buzz AI relay (server) puts you in control of your team’s data, messages, and files. Instead of relying on Block’s hosted servers, you run your own. The desktop app you download is just a client (window); the relay is where everything is stored.
To start, sign up for the Buzz desktop app and create an account. Click your account, hover over the public key, and copy the npub—that’s your Buzz identity. Next, go to Railway (or similar hosting), find the Buzz deploy box, and click configure. Paste your npub into the field, then hit deploy. Railway handles the complex setup—database, Redis, storage, and the Buzz app—connecting everything in under 30 seconds. You’ll get a URL for your relay; click it to see your fresh, empty community, then use the “Open in Buzz” link to join it through the app.
You can also add a custom domain: select it in settings, type your domain, pick a port, and add it. Buzz gives you DNS records to configure with Cloudflare or your domain host.
A key feature: you can enable “share compute” to let other members run agents on your machine, using your local open-source models. Or, push code to remote repositories hosted on your relay, integrated with GitHub. By self-hosting, you control the door—no one else sees your data unless you allow it.
Lesson 3: Best practices and pitfalls
Self-hosting an AI relay (a server that stores your messages, files, and membership list) gives you control, but it comes with trade-offs. If you let Block host your relay, it works like Slack but with more data control. If you self-host, you own everything—but you also own the bill.
Your Buzz desktop app is just a client (a window into a server). The relay is the actual server holding your data. When you self-host, you paste one public key (your npub, or Buzz identity) into a deploy button, and the system spins up the database, Redis, storage, and the Buzz app itself in under 30 seconds. You can add a custom domain by copying DNS records to Cloudflare.
The biggest mistake: treating a self-hosted relay like an open playground. On your own relay, you must trust everyone you let in—they can upload files, generate content, and spend your compute resources. One creator runs three relays: a free community on Block's servers (an open playground where strangers can walk in), a business relay for clients and team huddles that he self-hosts, and a third for personal projects. His rule: if it's public or experimental, let Block host it. If it's business, clients, or paying community, host it yourself.
Also, the door isn't shut by default when you set up a relay—new members can join unless you lock it down. For business relays, restrict access immediately. Best practice: use Block's hosted relay for open communities, self-host for anything private, and always check your resource usage since you're paying for what members consume.