Self-hosting

Run Call Copilot on your own infrastructure. Your data, your keys, your network.

Self-hosting is available on the Team Self-Hosted plan ($200/seat/month, or one-time license — contact us). You get the same source code we run, your own deployment, and Slack support.

Why self-host

Compliance. If your industry requires that call data stay on infrastructure you control (legal, healthcare, finance, defense, regulated EU), self-host is the only acceptable option.
Data sovereignty. Pick the region. The product has no cloud dependencies you can't replace.
Cost at scale. When you're past 20-30 seats, paying us monthly costs more than paying your own ops team. Self-host with a one-time license is cheaper.
BYO models. Use your existing OpenAI / Azure OpenAI / Anthropic / Bedrock / on-prem Llama infrastructure. You already paid for those tokens.

What you need

A Linux server (anything from 1 vCPU / 1 GB RAM upward; t3.medium / 2 GB recommended)
Node.js 20+
ffmpeg (for audio conversion)
API keys for one transcription provider (Groq, OpenAI, etc.) and one LLM provider
A domain name (recommended) and TLS cert (Let's Encrypt works)
Optional: Docker, Kubernetes, an existing reverse proxy

Quick start (Docker Compose)

The simplest deployment, single host, no orchestrator:

# 1. Clone the repo
git clone https://github.com/meta-slayer/realtime-call-copilot.git
cd realtime-call-copilot

# 2. Create your .env
cat > .env <<'EOF'
NODE_ENV=production
PORT=3000

# Auth
JWT_SECRET=$(openssl rand -hex 32)
# Optional legacy single-user (skip if using multi-user signups)
# AUTH_EMAIL=you@example.com
# AUTH_HASH=$(node -e "console.log(require('bcryptjs').hashSync('your-password', 10))")

# Pick providers
TRANSCRIPTION_PROVIDER=groq
GROQ_API_KEY=gsk_...
LLM_PROVIDER=openai
OPENAI_API_KEY=sk-...

# Optional: Stripe (skip for free internal use)
# STRIPE_SECRET_KEY=sk_live_...
# STRIPE_WEBHOOK_SECRET=whsec_...
# STRIPE_PRICE_SOLO=price_...
# STRIPE_PRICE_TEAM=price_...
EOF

# 3. Build + run
docker compose up -d

# 4. Visit http://localhost:3000

Bring-your-own LLM provider

The product supports a pluggable provider abstraction. Out of the box:

groq — Whisper-large-v3 (transcription) + Llama 3.x (LLM)
openai — Whisper-1 + GPT-4o-mini / GPT-4o
mistral — Mistral Large / Mistral Small
local-whisper — local Whisper.cpp (transcription only)

Adding your own provider: implement transcribe(buffer, mime, opts) or generate(messages, opts) in server/providers.js and register it. See the existing implementations for a template.

Production hardening checklist

☐ Set JWT_SECRET to a long random string (32+ bytes)
☐ Run behind nginx / Caddy / Cloudflare with TLS (Let's Encrypt is fine)
☐ Configure SMTP env vars for transactional emails
☐ Put the server behind a firewall — only 443 (or your reverse proxy) public
☐ Back up data/ directory (contains calls.json, clients.json, users.json, audio files)
☐ Monitor disk usage (audio files can grow fast on heavy use)
☐ Set up log rotation
☐ If multi-tenant, configure usage caps per plan in server/usage.js
☐ If you want auto-updates for the desktop client, host /downloads/ and bundle the right URL in electron/main.js

Kubernetes / Helm

A reference Helm chart is in the repo at deploy/helm/ (in progress for the next release). Stateless server pod + persistent volume for audio + ingress with TLS. Reach out for production deployment review.

Updates

Self-hosted licensees get update notifications and access to the same releases as the hosted product. git pull && docker compose up -d --build is the upgrade path.

Support

Self-hosted plan customers get:

Dedicated Slack channel
Deployment review (we'll look at your config before you go live)
Email response within 1 business day
Quarterly check-in calls

License

The source code on GitHub is available for inspection and personal use. Production deployment for an organization requires a Self-Hosted license ($200/seat/month subscription or one-time license — contact us for pricing). This is the agreement that funds ongoing development.

Get started

Email hakimnmo@gmail.com with: your team size, the regulatory regime that applies (if any), and which LLM provider you'd prefer to bring. We'll send a contract within 48 hours.