A complete browser-based AI development environment you install on any VPS. Agent chat, sandboxed code execution, remote desktop, project management — connect your own AI providers and own your entire stack.
Ubuntu 22.04/24.04 or Debian 12+. Takes 2–5 minutes. Disconnected? Run it again — it resumes.
A $20/mo VPS with flat-rate AI subscriptions costs a fraction of dedicated hardware + pay-per-token API keys. The portal is free — here's what the total setup looks like.
* API cost estimates based on moderate-to-heavy developer usage. OAuth flat-rate subscriptions cap your monthly cost regardless of usage volume.
| Capability | BridgesLLM Portal | Mac Mini + APIs | Gaming PC + APIs | Cloud IDEs |
|---|---|---|---|---|
| Access from any browser | ✓ Built-in | ~ VNC/RDP required | ~ VNC/RDP required | ✓ Yes |
| Flat-rate AI model costs | ✓ OAuth subs | ✗ Pay-per-token | ✗ Pay-per-token | ✗ Pay-per-token |
| Claude / Codex / Gemini | ✓ 6+ providers | ~ DIY setup | ~ DIY setup | ~ Limited |
| OpenClaw agent framework | ✓ Built-in | ✗ Self-managed | ✗ Self-managed | ✗ No |
| Sandboxed code execution | ✓ Isolated per-project | ~ Manual setup | ~ Manual setup | ✓ Yes |
| Remote desktop (browser) | ✓ NoVNC built-in | ✗ Extra software | ✗ Extra software | ✗ No |
| Project management + AI | ✓ Included | ✗ Not included | ✗ Not included | ✗ No |
| Self-hosted email server | ✓ Included | ✗ No | ✗ No | ✗ No |
| Self-updating dashboard | ✓ Auto-updates | ✗ Manual | ✗ Manual | ✓ Managed |
| Setup wizard (no CLI needed) | ✓ Browser-based | ✗ No | ✗ No | ✓ Yes |
| Works on $20/mo hardware | ✓ Yes | ✗ $800+ upfront | ✗ $1,200+ upfront | ~ Usage-based |
| Your data on your server | ✓ Always | ✓ Local | ✓ Local | ✗ Cloud-hosted |
Every tool you need to vibe code, run agents, and ship projects — without paying for or managing them individually.
BridgesLLM Portal runs on OpenClaw — the open-source agent framework that connects to every major AI provider via OAuth. Flat-rate access, no token counting, no surprise bills.
OpenClaw (agent framework)
Claude Code (Anthropic)
Codex (OpenAI)
Gemini (Google)
Ollama (local models)OpenClaw, Claude, Codex, Gemini, and Ollama names and logos are trademarks of their respective owners and are shown for identification only.
Get a VPS, run one command, and follow the steps below. No CLI experience required.
Hetzner, DigitalOcean, Vultr, Linode — any Ubuntu 22+ server. 2GB RAM, $5–40/mo.
One curl command. Takes 2–5 minutes. Prints your portal URL when done.
Domain, SSL, admin account — all in the browser. No config files to touch.
Run openclaw onboard in the portal terminal. Start chatting immediately.
Everything you need for each step, including commands and tips.
SSH into your VPS and paste the install command. The script installs all dependencies, sets up the database, configures SSL, and starts the portal.
The installer is interactive — it will ask for your domain name and create your admin account. When it finishes, it prints a URL to access your portal.
Open the portal URL in your browser. The setup wizard walks you through domain configuration, SSL certificates, and your admin account. Everything is handled in the browser — no config files to edit.
Navigate to the Terminal section in your portal and run the OpenClaw onboard command. This connects your AI provider API keys (Anthropic, OpenAI, Google, etc.).
Follow the prompts to enter your API key. When asked about existing values, select "Keep existing" to preserve the gateway configuration the installer set up.
openclaw onboard any time to add or change API keys. The portal will automatically reconnect.Go to Agent Chats in the sidebar. Select your AI provider and send a test message. If the dashboard shows a green status indicator, your agent is connected and ready.
That's it — your AI workstation is live. Start chatting, create projects, or explore the file manager and remote desktop.
Copy the prompt below and paste it into your OpenClaw agent's workspace file (e.g. AGENTS.md or system prompt). This teaches the agent how your portal works — where files live, how projects are structured, and how to interact with users through the portal interface.
↓ Scroll down to the "Agent Portal Prompt" section to copy it.
The dashboard may show green while the real-time connection is still establishing. Click the Reconnect button on the dashboard, or restart the portal service:
Make sure your domain's DNS A record points to your VPS IP. After DNS propagates, re-run the installer to pick up the domain and provision SSL:
The installer detects the existing installation and only updates what's needed.
The portal includes a built-in self-update mechanism. From your portal's Settings → System, click Check for Updates. Or from the terminal:
If something isn't working, check that both services are running:
Portal and gateway logs are available via journalctl:
Run the onboard command again from the portal's Terminal section. Select "Keep existing" for values you don't want to change:
Paste this into your OpenClaw agent's workspace instructions to teach it how your portal works. Your agent will know where to find files, how projects are structured, and how to help users through the interface.
## BridgesLLM Portal Integration
You are an AI assistant running through the BridgesLLM Portal — a self-hosted web-based
AI workstation. Your human interacts with you through a browser interface that includes
agent chat, a file manager, a web terminal, project workspaces, and a remote desktop.
### Portal Architecture
- **Agent Chat**: Your primary interface. Messages stream in real-time via WebSocket.
The chat supports markdown, code blocks with syntax highlighting, and inline tool calls.
- **File Manager**: Users can upload, browse, and organize files at `/portal/files/{userId}/`.
When a user says "I uploaded a file," check that directory.
- **Projects**: Each project lives at `/portal/projects/{userId}/{projectName}/`.
Project agents are sandboxed to their project directory.
- **Terminal**: A full browser-based terminal (xterm.js). Users may run commands here
while chatting with you. You share the same server environment.
- **Remote Desktop**: noVNC-powered browser desktop for GUI applications.
### File Locations
- User uploads: `/portal/files/{userId}/`
- Project files: `/portal/projects/{userId}/{projectName}/`
- Portal backend: `/opt/bridgesllm/portal/backend/`
- Portal frontend: `/opt/bridgesllm/portal/frontend/`
- OpenClaw config: `~/.openclaw/openclaw.json`
- OpenClaw workspace: `~/.openclaw/workspace-main/`
### Working with the Portal
1. When users mention files they uploaded, check `/portal/files/` first.
2. When working in a project context, stay within the project directory.
3. The portal handles streaming — write complete responses; don't worry about chunking.
4. Code blocks render with syntax highlighting. Use language tags (```python, ```bash, etc.).
5. The user can see your tool calls in the chat interface. Be transparent about what you do.
6. If you need to install packages or run long processes, note that the terminal is shared.
### Useful Commands
- `openclaw status` — check gateway health
- `openclaw onboard` — reconfigure API keys
- `systemctl status bridgesllm-product` — check portal service
- `systemctl status openclaw-gateway` — check gateway service
### Boundaries
- Stay within your designated workspace or project directory.
- Do not modify portal source files unless explicitly asked.
- Do not access other users' files or projects.
- Ask before running destructive commands (rm, drop, etc.).
If you can SSH into it, you can run BridgesLLM Portal on it.
Free to install. Free to run. Pay only for the VPS and the model subscriptions you want.