What it captures
Leads, phone numbers, emails, urgency, postcode, appointment requests, transcripts, inbound email enquiries, staff handoff summaries and a searchable event history.
Open-source AI receptionist
Run a private front desk agent on a normal CPU server.
FrontDeskAgent is now available as a self-hosted project for teams that want local control, easy setup, website-trained answers, phone/SMS workflows, email intake, booking handoffs, CRM automations, calendar visibility, and practical business automation without needing a GPU first.
CPU
SQLite, Flask, setup wizard, Twilio, Telnyx, SMTP, CRM webhooks, calendar feed, OpenZero-first auto routing, Ollama, llama.cpp, GGUF and a no-model fallback.
How it runs
Start with the setup wizard. The default auto route tries OpenZero, Ollama, llama.cpp, optional hosted API fallback, and then the built-in no-model fallback.
Real integrations
Use Twilio voice/SMS webhooks, Telnyx SMS, SMTP email, email parser hooks, outbound callback hooks, booking webhooks, CRM/n8n/Zapier/Make webhooks, and a token-protected calendar feed.
Website-trained
Paste public service, FAQ, price, policy, location, or script page URLs into the knowledge base and review the imported text before production use.
Who it helps
Plumbers, clinics, hotels, agencies, admissions teams, estate agents, construction firms, local services and small businesses with missed calls.
Why open source
Developers can inspect the stack, deploy it privately, adapt it to their own workflows, and still ask for managed setup when they want support.
Ready playbooks
The companion playbook library includes structured intake flows for common industries, with fields, urgency rules, SMS templates and handoff summaries.
Quick install
Clone, run the wizard, start the receptionist.
The first boot works without a model. The setup wizard generates safe secrets, admin login, webhook secret, calendar token, local-first AI routing defaults and the exact provider URLs to paste into Twilio, OpenZero, n8n, Zapier, Make or a CRM.
git clone https://github.com/ResearchForumOnline/FrontDeskAgent.git
cd FrontDeskAgent
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python -m frontdeskagent.setup_wizard
python -m frontdeskagent.appCPU-first model path
Start small, then upgrade only when the workload proves it.
Good first tests use OpenZero or compact instruction models through Ollama. Larger GGUF models can be added later through llama.cpp when the server has enough RAM and the use case needs more reasoning. Hosted APIs stay optional.
Auto route
Use the default local-first route order: OpenZero, Ollama, llama.cpp, optional hosted API, then rules fallback.
Ollama
Use compact CPU-friendly models such as small Qwen, Llama or Gemma variants.
OpenZero
Let OpenZero route work, supervise events, and connect the receptionist to wider local automation.
Managed or self-hosted
Use the open-source repo yourself, or request a production build.
The public repo is for self-hosted installs. FrontDeskAgent Online also offers managed planning, hosting, setup, call-flow design, tuning, security review and business integrations.