Install Talk To My Agent
The voice gateway installs on your own server with a single command. No infrastructure to rent, no Docker to configure - just paste and go on any supported platform below.
Quick start
After you provision a phone number in the voice portal, the checkout screen gives you a one-line install command with a signed token. It looks like this:
curl -sSL https://voice.moltbot.ninja/install.sh | bash -s -- --token YOUR_TOKEN --accept-licenseThe installer verifies the download against a pinned SHA-256, presents the license for your approval, and sets the gateway up as a systemd service. The whole thing takes about a minute on a modern Linux box.
Supported platforms
Updated regularly - check back if your platform isn't ready yet.
Linux - x86_64 (64-bit)
Ready to installUbuntu 20.04+, Debian 11+, Rocky / Alma / RHEL 8+, Fedora, Arch, WSL2
Linux - arm64 / aarch64
Ready to installRaspberry Pi OS 64-bit, Ubuntu arm64 servers, AWS Graviton, Oracle Ampere
macOS - Apple Silicon (M1 - M4)
Coming soonMacBook Air / Pro M1 or later, Mac mini M-series, Mac Studio
Binary signing + Apple notarization in progress. Target: next release.
macOS - Intel (x86_64)
Coming soonMacBook Pro / Air pre-2020, Intel Mac mini, iMac
Ships alongside the Apple Silicon build. Target: next release.
Linux - armv7 (32-bit)
ExperimentalRaspberry Pi 2/3/4 running 32-bit Pi OS
Builds are published but not yet fully validated on real hardware. Use at your own risk; please report any issues.
Windows and NixOS
We don't build native Windows or NixOS binaries and have no plans to. On Windows, install WSL2 (Ubuntu works great) and follow the Linux x86_64 path - it's a 5-minute setup and everything just works.
System requirements
1 GB RAM minimum (2 GB recommended). ~200 MB disk for the binary and its runtime files.
Any modern 64-bit CPU. A small cloud VPS or a Raspberry Pi 4 is more than enough for dozens of concurrent calls.
Outbound HTTPS only. No inbound ports to open - traffic flows through an encrypted Cloudflare tunnel.
Installer runs as a regular user with sudo to install systemd unit. The gateway itself runs sandboxed, never as root.
Linux compatibility note
Our Linux binaries are built against glibc 2.28, which is included in every major distribution released after 2019 (Debian 10+, Ubuntu 20.04+, RHEL/Rocky/Alma 8+). If you're on a very old LTS (Ubuntu 18.04, CentOS 7), the installer will tell you upfront and ask you to upgrade.
Common questions
Do I need Docker?
No. The gateway ships as a single self-contained binary. No Docker, no Node.js install, no npm dependencies - just one file + a systemd unit.
Can I install on a shared server or VPS?
Yes, but a small dedicated VPS is best (DigitalOcean, Hetzner, Linode, AWS Lightsail all work). The gateway is lightweight but expects to own port 3335 on localhost. It does not need a public IP - all traffic is outbound through a Cloudflare tunnel.
What happens if my server reboots?
The gateway is installed as a systemd service with Restart=always. It comes back up automatically and reconnects to the tunnel. Calls that were in progress drop (as they would with any reboot), but inbound callers see a normal ring-no-answer and can call back immediately.
How do I update the gateway?
Updates ship automatically - the gateway verifies its own binary hash on startup and fetches the latest version when a new release is published. There's nothing for you to do.
Can I run it on a Raspberry Pi?
Yes - Raspberry Pi OS 64-bit (arm64) is fully supported. A Pi 4 with 2 GB of RAM handles real production traffic comfortably. The 32-bit version of Pi OS is still experimental; if you have a choice, use 64-bit.
I'm on macOS. When can I install?
Soon. Apple requires every distributed binary to be code-signed and notarized through their developer program. We're in the enrolment + setup phase now. The moment it's ready, the macOS command appears on your checkout screen - no action needed on your side.