Pricing
We build it. We prove it. We keep it alive. Pricing is based on the size of your codebase, not the number of repos.
Base Rate
of mapped source code
~1M tokens ≈ 50–100K lines of code, depending on language
Proof Pilot
mapped source code + onboarding & eval
- Intent Layer built and delivered
- Eval run on your actual tasks
- Benchmark report: with vs. without
- Delivered in 48 hours
- You keep the Intent Layer either way
Money-back guarantee
No improvement on your eval, no charge.
Expansion
additional source code mapped
- Roll the Intent Layer across more code
- Patterns transfer — efficiency improves
- Small repos bundled cost-effectively
- Same eval methodology on each
- Cumulative benchmark across your org
Typically follows a successful Proof Pilot.
Discuss ExpansionFreshness Retainer
ongoing maintenance
- Automated updates on every commit/merge
- Zero maintenance effort from your team
- Staleness detection and alerts
- Context improves every sprint
- Priority support if anything breaks
We manage the automation. Zero engineer time from your team. This is where it compounds.
Discuss RetainerHow it scales
Why tokens, not repos?
Many companies have dozens of small repos (50K–100K tokens each). Pricing by token means you don't pay $10K per tiny repo — bundle them together and pay for what's actually mapped.
What counts as a token?
Source code in your repository. We measure the codebase before starting. Tests, docs, and config files that aren't part of the mapping don't count. You can measure your own codebase with our free open-source tool: go install github.com/Napageneral/tokcount@latest && tokcount .
The Intent Layer needs scale to shine
The eval shows the biggest improvements on codebases above 1M tokens — that's where agents start struggling without structured context. Below that, agents can often manage on their own.
Retainer pricing
Monthly pricing depends on total tokens under maintenance and how frequently your codebase changes. Retainer keeps the Intent Layer current and compounding.
Common Questions
How is "improvement" measured?
We take real engineering tasks from your backlog, run them through AI agents with and without the Intent Layer, and score the output on correctness, completeness, and quality using an LLM judge. The methodology is agreed upon upfront before any work begins.
What do I actually get from the Proof Pilot?
A set of structured context files (.md) that live in your repository, a benchmark report showing agent performance with and without the Intent Layer, and a walkthrough of the results. You keep all deliverables regardless of the outcome.
Can we try building the Intent Layer ourselves?
Absolutely. Our blog has a full technical breakdown of the methodology. Some teams try it themselves first and then engage us when they see the quality gap — we have tooling and a refined process that produces stronger results faster.
What kind of repo access do you need?
We need access to the repos being mapped. Code is processed through SOC2-certified model providers (Anthropic, OpenAI) that don’t train on API data. We can use your API keys and preferred provider, or run the cartographer entirely within your infrastructure.
How long before we see ROI?
The Proof Pilot delivers measurable results in 48 hours. If you move to a retainer, the system improves continuously — every code change feeds back into better context.
What if we have lots of small repos?
Token-based pricing means small repos get bundled together. If you have 20 repos at 50K tokens each, that's 1M tokens total — one unit of pricing, not 20 separate engagements.
Start with a Proof Pilot.
48 hours from repo access to proof. No improvement, no charge.
You keep the Intent Layer either way.
Book a 30-Minute Call