Self-hosted AI receptionist that runs on your own servers
git clone https://github.com/lobbystack/lobbystack.gitMIT-licensed and built to run with Docker Compose. Recordings, transcripts, and customer records stay in your own database, and calls run on your own Twilio and OpenAI accounts.
MIT License with no license fee
Deploys with Docker Compose or the Railway template
Runs the same code as LobbyStack Cloud
How a call moves through your server
Everything inside the dashed line runs on your server. Twilio and OpenAI run on your own accounts.
Caller
Dials your business number
Twilio + OpenAI
GPT-Live answers the call from your accounts
Your serverdocker compose
Dashboard
Admin app and API
Worker
Runs calls and background jobs
PostgreSQL
Calls, contacts, settings
Redis
Queues and rate limits
Storage
Recordings and documents
Recordings, transcripts, and customer records stay in the bottom row.
Optional
Google Calendar, email, analytics, and website import
Deploy with Docker Compose
The official setup runs on one server. On Railway, you can deploy the LobbyStack template instead, and it generates the secrets for you.
- 1
Clone the repository
git clone https://github.com/lobbystack/lobbystack.git && cd lobbystack - 2
Generate secrets
The script creates .env and fills every database password and internal secret with a random value.
node scripts/generate-env.mjs - 3
Add your provider keys
Open .env and set APP_BASE_URL, OPENAI_API_KEY, and your Twilio credentials.
- 4
Start the stack
Migrations run before the dashboard and worker start.
docker compose --env-file .env up -d --build
Choose where call data lives
Recordings, transcripts, contacts, and settings stay in the PostgreSQL database and storage you run. You set the retention, backup, and deletion rules. Twilio and OpenAI still process live call audio under your accounts, so review their terms for your use case.
- Separate database roles and row-level security for each service
- Recordings on a local volume or any S3-compatible bucket
- Backups and restores on your schedule
Change the code to fit your workflow
You get the TypeScript monorepo that runs LobbyStack Cloud. Edit the prompts, intake questions, and call rules, or connect LobbyStack to internal systems your team already uses.
- System prompts live in the packages/agent-core workspace
- Point chat and embeddings at any OpenAI-compatible endpoint
- Pin a release and upgrade after you've tested it
Know what you'll pay before you deploy
LobbyStack charges no license fee for self-hosting. You pay your hosting provider, Twilio for numbers and call minutes, and OpenAI for Realtime usage, each on your own account. Count the hours your team will spend on updates, backups, and monitoring before you compare it with a Cloud plan.
- Twilio and OpenAI bill you directly
- No per-seat or per-minute fee from LobbyStack
- Cloud plans start free if you'd rather not run servers
LobbyStack Cloud or self-hosted
Both run the same receptionist. Pick self-hosting when you need the data on your own servers and have someone to run them.
| Topic | Self-hosted | LobbyStack Cloud |
|---|---|---|
| Who runs the servers | You, on Docker Compose, Railway, or your own cloud | LobbyStack |
| Software cost | No license fee under the MIT License | Free plan, then $30 or $100 a month with minutes included |
| Phone and AI usage | Billed to your own Twilio and OpenAI accounts | Included in your plan's minutes |
| Where call data lives | Your PostgreSQL database and storage | LobbyStack's managed infrastructure |
| Updates | You pull a release and redeploy when you're ready | Roll out automatically |
| Backups and monitoring | Your team sets them up and tests restores | Handled for you |
| Code | The same core, and you can change it | The open-source core |
Rather not run servers? Start on LobbyStack Cloud for free.
Questions about self-hosted AI receptionists
What is a self-hosted AI receptionist?
A self-hosted AI receptionist runs on your own servers or cloud infrastructure instead of a third-party SaaS platform. You control the data, the model, the deployment environment, and the integrations. LobbyStack is open source and supports self-hosted deployments for teams that need full control.
Is LobbyStack open source?
Yes. LobbyStack uses the MIT License. You can inspect, fork, modify, distribute, and deploy the source code under that license. Keep the copyright and permission notices with copies or substantial portions of the software.
What are the self-hosting requirements?
The Docker Compose guide asks for Docker Engine 24 or later with Compose v2, Node.js 22 or later to generate secrets, and a server with at least 2 vCPU, 4 GB of RAM, and persistent disk. Live calls also need a domain with HTTPS for the dashboard, plus Twilio and OpenAI accounts. The Railway template sets up the services and databases for you.
Can I use my own LLM or API key?
You use your own API keys. Voice calls run on OpenAI GPT-Live with your OpenAI account. Text generation and knowledge embeddings accept any OpenAI-compatible endpoint, so you can point them at another provider or a model you host.
Is self-hosting suitable for agencies and resellers?
Yes. The MIT License permits agencies to modify and distribute LobbyStack for client work. The LobbyStack name, logos, and branding remain subject to separate trademark rights.
How do updates work for self-hosted deployments?
You pull changes from the GitHub repository, review the changelog and migration notes, and redeploy through your own release process. Pin the version you have tested instead of automatically deploying every upstream change.
What about data privacy and compliance?
Self-hosting gives you control over LobbyStack's application deployment and stored business data. Calls can still be processed by configured telephony, AI, hosting, and integration providers, so review each provider's data handling and complete your own privacy and compliance assessment.
Do you offer support for self-hosted installations?
Start with the repository documentation and public GitHub issue tracker. For questions that do not belong in a public issue, contact the LobbyStack team through the support address listed on the site.
Can I customize the voice, prompts, and behavior?
Yes. Self-hosted deployments give you full access to the prompt templates, voice settings, greeting scripts, and routing rules. You can tailor every aspect of the caller experience.
How does self-hosted pricing work?
The MIT-licensed source code has no separate software license fee. You remain responsible for infrastructure plus any telephony, AI, storage, monitoring, and integration-provider charges used by your deployment.
Deploy LobbyStack on your own servers
The self-hosting guide covers the services, provider accounts, and backups you'll run. The Docker Compose guide walks through a single-server deployment.
No credit card required. Cancel anytime.