Show HN: Built Valmi to bill AI Agents(Open Alternative to Paid.ai, Legacy Zuora (github.com)

4 points by rajvarkala ↗ HN
Hi HN,

I am Raj, founder of Valmi.

We were building AI agents and quickly hit a pricing problem.

Tokens, API calls, and seats made no sense for agents. One agent run might be cheap and useless; another might be expensive and replace hours of human work. Billing on usage felt disconnected from real value.

What customers actually asked was: “Did it finish the task?” “Did it resolve the ticket?”

That pushed us toward outcome-based billing - charge only when the agent delivers something meaningful.

But existing billing systems don’t handle this well:

outcomes aren’t first-class concepts costs vary per run (LLMs, tools, retries) it’s hard to see margins per agent or customer

So we built Valmi:

1. outcomes as billable units 2. cost + margin tracked per agent run 3. outcome, usage, or hybrid pricing 4. open-source SDKs, self-hostable stack

Outcome billing also builds trust: customers pay for results, not experiments.

Available on GitHub (https://github.com/valmi-io/value).

I also wrote my hypothesis here: https://www.valmi.io/blog/an-imperative-for-ai-agents-outcom...

Would love your feedback! Also hosted at https://value.valmi.io

-Raj

5 comments

[ 3.2 ms ] story [ 21.5 ms ] thread
I am curious to know your opinions to the questions I am thinking about.

What do you think are the challenges with AI Agents pricing? How you are currently pricing your AI Agents work?

Would be great to know what HN community thinks.

Hi Raj, If the cost of running an AI agent can vary a lot from task to task, how do teams determine pricing while keeping profits and financial risk under control?
CPA here. The math on this is brutal because of the fixed fee floor. If an agent performs a $0.05 task, a standard $0.30 + 2.9% fee structure puts you at a negative margin immediately. This is a solvency issue, not just a pricing one. You literally cannot monetize granular, high frequency agent traffic with legacy banking rails. The only way the P&L works for micro services ($0.01 to $0.10) is via rails where the settlement cost is negligible (<$0.001), regardless of the asset class used.
Who owns the definition of outcome? Furthermore, how do we achieve alignment when the desired results are nuanced or subjective?
Hi Raj, How do you programmatically and objectively define a successful outcome in a way that prevents billing disputes between the provider and their customers?