Your AI.
Your rules.

Run AI agents on your own server. Bring your own API keys or run open-source models on dedicated GPUs. Connect to WhatsApp, Slack, and Telegram. One binary. No DevOps.

Or self-host: curl -fsSL https://getsolon.dev/install.sh | sh

Models

Can I run this?

Open-source models, secured by Solon. See what runs on your hardware.

Run on your machine
Free — install Solon and go
?
Phi-3 Mini 2b
1.7 GB
?
Qwen 2.5 1.5b
1.3 GB
?
DeepSeek R1 1.5b
1.1 GB
?
Llama 3.2 3b
2 GB
?
Gemma 3 4b
3.3 GB
?
Mistral 7b
4.1 GB
?
Llama 3.1 8b
4.7 GB
?
Qwen 2.5 7b
4.7 GB
?
DeepSeek R1 7b
4.7 GB
?
Gemma 3 9b
5.5 GB
?
Phi-4 14b
9.1 GB
?
DeepSeek R1 14b
9 GB
?
Gemma 3 27b
17.2 GB
?
Qwen 2.5 32b
20.5 GB
?
DeepSeek R1 32b
20 GB
curl -fsSL https://getsolon.dev/install.sh | sh
Too large for your machine?
Deploy on managed GPU
NVIDIA A100 / H100 / H200 in Europe
DeepSeek R1 70b
43 GB
Llama 3.1 70b
40 GB
Command R+ 104b
60 GB
Llama 3.1 405b
230 GB
Deploy on GPU — from $3.49/hr View all plans
Agents

Agents that do real work

Not another chatbot. Solon agents run autonomously — connected to your tools, your channels, and your data.

Autonomous Agents

Deploy agents with tools, skills, and MCP server connections. They research, write, analyze, and execute tasks on their own — not just respond to prompts.

Channel Integrations

Connect agents to WhatsApp, Telegram, Slack, and Discord from the dashboard. Your agents meet your users where they already are.

Tiered Security

Control what each agent can do. Tier 1: inference only. Tier 2: tools and internet. Tier 3: persistent storage. Tier 4: full capabilities with MCP and custom skills.

Models

Your models. Your API keys.

Use any model from any provider — or run your own. One unified API with auth, rate limiting, and analytics built in.

Bring Your Own Keys

Use your existing Anthropic, OpenAI, or NVIDIA API keys. Solon proxies requests with auth, rate limiting, and usage tracking — one API for all your providers.

Run Open-Source Models

Pull and run Llama, Gemma, Qwen, Mistral, and more on dedicated NVIDIA GPUs. Or run locally on your Mac with llama.cpp and Metal acceleration.

OpenAI-Compatible API

Drop-in replacement for the OpenAI API. Any tool, SDK, or framework that works with OpenAI works with Solon. Zero code changes.

Infrastructure

Your server. Your data.

Dedicated hardware, mandatory security, and full data sovereignty. We manage the server — you own everything on it.

Dedicated Hardware

Every managed instance runs on its own server. No shared compute, no noisy neighbors. Full tenant isolation with automatic TLS.

Mandatory Auth

Every request requires an API key. Keys are bcrypt-hashed, never stored in plaintext. There is no --no-auth flag. Security is the default, not an option.

Open Source

Solon is MIT-licensed. Run it yourself for free forever, or let us manage the infrastructure so you can focus on building.

Solon vs. doing it yourself

Your AI, your rules — without the infrastructure headache. Get everything set up in minutes instead of months.

Task Solon DIY
Provision a server 5 minutes 2-4 hours
Install AI runtime + auth + TLS Included 1-2 days
Deploy agents with tools Pre-configured 1-2 weeks
Connect WhatsApp/Slack/Telegram Dashboard toggle 1-2 weeks
Deploy open-source model One click 4-8 hours
Add API keys (Anthropic, OpenAI) Dashboard Custom integration
Security hardening + monitoring Included 1-2 days
Ongoing maintenance $0 extra $50-100K/yr engineer time
175,000+

Ollama instances exposed to the internet without authentication

The current default path for self-hosted AI is a security disaster. Solon makes auth mandatory — there is no --no-auth flag.

Start on your terms.

Deploy a managed instance in minutes, or self-host on your own hardware. Either way, you own it.

Prefer to self-host?

curl -fsSL https://getsolon.dev/install.sh | sh