Micropayments
Why Subscriptions Don’t Work for AIs?
At Kirha, we believe subscription models are fundamentally incompatible with the operational and architectural needs of AI agents. The issues fall into three categories:
1. Economics
AI inference is a commodity, like electricity or gasoline, and selling it via subscription doesn’t scale. Power users overwhelm the system and break the economics.
On the other hand, smart reasoning in AI requires granular, composable data. Flat-rate subscriptions per dataset or provider are inefficient when agents need to dynamically access many data points across domains.
Paying for bulk access to data that may not be used results in wasted cost and poor scalability.
2. Operations
Managing multiple data providers introduces overhead:
- API keys rotation
- Secrets distribution
- Rate and spend limits
Doing this repeatedly across many providers adds friction and increases security risk.
3. Reliability
We’re obsessive about data provenance—not just because it builds confidence in AI outputs, but because it also enables error tracing, allowing bugs or inconsistencies to be identified and resolved. This directly enhances the reliability of AI systems.
For all these reasons we’re building infrastructure for AI agents to acquire data at the level of individual data points, with end-to-end traceability. Each acquisition generates a proof, stored immutably on a public blockchain, minimizing trust assumptions and enabling full auditability.
In our first collaboration with Swiss 6022, Kirha will be paid in their native token on Polygon, creating an indelible, onchain record of data acquisition events.
This unlocks verifiable transparency in how AI systems consume and pay for information.
However, while onchain payments are transparent, they’re not fast or cheap enough yet—transaction fees can reach $0.30 for a single USDC transfer in times of network congestion, and confirmation delays slow down AI response times.
Money just doesn’t move as fast as information.
That’s why we let users prepay in fiat (via one-time payments or subscriptions) to acquire Kirha credits. Regardless of the way you pay in Kirha, pay per use, top ups, subscriptions, you will always be running micropayments for each data obtained.
Each AI prompt includes a Kirha receipt, detailing the planning process: which data points were considered, selected, and why.
We’re actively R&D’ing the cheapest, most scalable way to publish these receipts onchain, to maintain auditability without compromising performance.
Don’t trust, verify.