{"id":46942551,"title":"Show HN: Molinar – Open-source alternative to ai.com (AGPL-3.0)","points":1,"user":"novelica","time":1770622160,"time_ago":"7 hours ago","type":"link","content":"

Hey HN,

I built a managed platform for OpenClaw (the open-source AI agent framework) and shipped the whole thing in a day. Then I open-sourced the platform itself.

The problem: Running your own AI agent means a Mac Mini in your closet, praying your wifi holds, and becoming a part-time sysadmin. Most people give up before they start.

The solution: 3 steps, 5 minutes, done.

1. Sign up\n2. Paste your Anthropic API key + Telegram bot token\n3. Hit launch

You watch your agent boot in real-time. When it hits "Ready," it's live on Telegram - running 24/7 without your laptop open.

What actually happens when you hit Launch:

- ECS Fargate spins up an isolated container (FARGATE_SPOT, 2 vCPU/4 GB)\n- Your API keys are pulled from AWS SSM Parameter Store (SecureString), encrypted at rest and never stored in our database\n- Each container gets its own ENI with an egress-only security group - no inbound ports, the agent initiates all connections\n- A background process patches the OpenClaw config for Telegram DM access\n- CloudWatch logs stream back to the dashboard, parsed into setup phases: provisioning -> configuring -> health check -> nginx -> gateway ready\n- Supabase Realtime pushes updates to the browser in real-time (3s polling during setup)

Stack: Next.js, Stytch B2B auth, Supabase (Postgres + Realtime), AWS ECS Fargate, SSM Parameter Store, CloudWatch, Stripe.

Security model:\n- Full container isolation — every agent is its own Fargate task, no shared compute\n- BYOK — you bring your own Anthropic key, we never see your usage or bill\n- Zero inbound ports — egress-only security group\n- Secrets encrypted in SSM with per-org/per-instance paths, never persisted to Supabase\n- You can inspect every line of code — the entire platform is AGPL-3.0

The agent itself can search the web, read/write files, run shell commands, manage cron jobs, monitor services, send messages - it's a full autonomous assistant. I run one myself that monitors my Sentry errors, checks my analytics, reads my Slack, and flags anything that needs attention. All through Telegram.

Pricing: $49/mo (starter) / $149/mo (pro). Or self-host the whole thing for free - it's all on GitHub.

I'm a solo developer. This is the entire platform, open source, built in a day on top of the existing OpenClaw OSS project.

Oh, and autoscale is on. ;)

business.molinar.ai\nGitHub: https://github.com/Molinar-AI/openclaw-business","url":"https://business.molinar.ai","domain":"business.molinar.ai","comments":[],"comments_count":0}