Pricing · USD
AI & LLM Integration
Assistants and AI features grounded in your own data — like Nova, in the corner of this site. Priced by how much it must know and how rigorously it’s evaluated.
Common bundles.
Presets built from the line items below. Most projects start at one of these and move as scope becomes clear.
Starter
$1,200 – $3k
A grounded assistant for one site or document set — Nova-shaped.
- Retrieval over your content
- One provider + fallback
- A can’t-invent-facts guardrail
- Low-cost hosting
Standard
$3.5k – $8k
A production assistant with memory, failover and evaluation.
- Hybrid retrieval + floor
- Multi-provider failover
- Conversation memory
- Evaluation harness
Custom
from $10k
Tool-using agents, fine-tuning, or AI woven through a product.
- Agents that take actions
- Fine-tuning where warranted
- Cost + abuse controls
- Ongoing evaluation
What makes up the price.
The line items
The price is the sum of the pieces you actually need. A row marked / unit is charged per that unit; the rest are once for the project.
- Retrieval setup $400 – $1.5k Crawling / ingesting your content and building the search index
- Grounding & guardrails $300 – $1k The scoring floor and prompt design that stop it inventing facts
- Provider chain & failover $200 – $700 / provider Each model provider wired in, with cost and outage handling
- Conversation memory $400 – $1,200 Follow-up questions that remember the thread, signed so it can’t be forged
- Evaluation harness $500 – $2k Automated checks that flag when answer quality drops
- Tools / actions $400 – $1.5k / action Letting it DO things (book, fetch, calculate), each action built and guarded
- Fine-tuning $800 – $4k A custom model where retrieval alone isn’t enough
- Deploy & cost controls $200 – $900 Hosting, rate-limiting, and abuse / spend guards