Custom AI Agent / Chatbot Build · From $4,500
Claude-powered agent built for your practice — voice or chat — for intake, FAQ, scheduling, or internal Q&A. Production-grade, HIPAA-aligned, deployed on infrastructure you control.
At a glance
What's included
A real, deployed agent on infrastructure you control — not a demo bot you have to host yourself afterwards.
- Discovery and use-case scoping with your team — what should the agent do, what should it NOT do
- Prompt engineering and persona design — voice, tone, escalation rules
- Integration with up to 2 systems (calendar plus EHR, CRM plus support desk, etc.)
- Single deployment surface (web embed, voice agent, or messaging channel — pick one)
- Knowledge-base ingestion of up to 50 documents (PDFs, Word, web pages) with retrieval tuning
- 30-day post-launch tuning window — we adjust prompts and exclusions based on real conversations
- Security review, audit logging, and (where applicable) BAA-aligned hosting setup
- Documentation: prompt library, escalation rules, conversation log access for your team
What's NOT included
Where the build ends.
- Multi-channel deployments — each additional channel after the first quoted separately ($1,500-$3,000)
- HIPAA BAA hosting fees from Anthropic, AWS, or other cloud providers (pass-through)
- Ongoing prompt maintenance after the 30-day tuning window (Automation Retainer covers this)
- Voice provider fees (ElevenLabs, Vapi, Twilio — pass-through)
- Net-new training data creation — we use what you already have
- Fine-tuning a custom model (rarely needed; we use prompt + retrieval instead)
How it works
- Buy via Stripe. $4,500 deposit (50%) processes immediately. Balance due at go-live.
- Discovery week. Use-case scoping, integration credentials, knowledge-base inventory, security review.
- Build (weeks 2-5). Prompt engineering, retrieval setup, integration wiring, audit logging. Weekly demo calls.
- UAT (week 6). Your team runs real scenarios. We tune persona, exclusions, and escalation rules.
- Go-live + 30-day tuning. Production deployment. Daily monitoring of conversations. Prompt iteration based on real interactions.
Frequently asked questions
Is it HIPAA-aligned?
Yes, with the right hosting choices. We deploy on AWS Bedrock or Anthropic enterprise tiers that support BAAs, encrypt all conversation data, enforce role-based access, and log every interaction. Standard consumer Claude does not have a BAA — we never deploy practices on that path.
Can I host it on my own infrastructure?
Yes. We support deployment to your AWS, Azure, or GCP tenant. Adds discovery time but no extra cost beyond infrastructure pass-through. If you have an existing security review process, we comply with it.
What models does it use?
Claude Sonnet by default (best balance of capability and cost) with Haiku for low-latency surfaces and Opus for complex reasoning paths. We pick per use case during discovery. We are not locked to one model — when newer Claude versions ship, we evaluate and migrate.
What if I want to expand to multiple channels?
Each additional deployment surface is roughly $1,500-$3,000 depending on complexity. Web embed plus voice agent plus SMS, for example, is three surfaces. We quote each separately so you do not pay for channels you are not using.
Does it remember conversations?
Yes — we configure session memory and, optionally, persistent user memory tied to your customer record. For HIPAA contexts we configure retention policies aligned with your record-keeping requirements.
Related products
Ready to start?
Buy the deposit and we'll have your discovery week scheduled within 5 business days. Or talk through use cases first — many practices think they need an agent when really a workflow automation does the job for less.
Buy now via Stripe → Book a consult