Kardinal opens a route optimization API designed to be driven by AI agents
- September 29, 2026
- 2 mins
Paris, September 29, 2026
Kardinal, a last-mile optimization specialist since 2015, today launched its route optimization API in self-serve.
While most optimization APIs were designed for human developers, Kardinal chose to rebuild its interfaces for the agent era, in order to automate business modeling and technical integration.
A technical challenge specific to coding agents
In February 2026, humans still accounted for 85% of technical documentation readers. By August, they were down to 33%, with AI agents now making up the rest (source: Mintlify*). Yet in route optimization, an agent relying on API documentation produces code that compiles and calls that succeed, but almost always formulates the wrong problem, one that cannot be executed in the field.
The gap between a real-world operational problem and its translation into an optimization model has always existed. Agents, non-deterministic by nature, amplify it: they quickly produce a plausible formulation, without the signal that would tell them it is wrong.
Interfaces redesigned to close this gap
Kardinal has invested in what the AI industry now calls harness engineering: designing how agents access and use software, so they get the right information at the moment they need it. Today, that means documentation written for agents, explaining when to use each setting and how to handle the hard cases of real-world logistics, along with tools built for the way they work, from a command-line interface to an SDK that catches mistakes early. Next, Kardinal will bring this guidance into the API itself: when a delivery plan doesn’t hold up, the response will explain why and what to try next, so the agent can keep reasoning instead of getting stuck.
The results show directly in usage: on the industrial cases Kardinal uses as benchmarks, the modeling failure rate drops from over 80% to under 5%.**
With this launch, we are making a bet: thanks to AI, route optimization will not just be faster, it will be more reliable. For ten years, we have refined our algorithms in contact with real operations, with all their exceptions and field constraints. Today, we are passing this same rigor on to agents: reasoning correctly about the real problem, rather than acting fast on a simplified version of it. This is what has always made the difference for our clients. It is now within reach of anyone building with agents.
Available today
- Developer console, API keys, usage tracking, Python SDK and documentation designed to be machine-readable: console.kardinal.ai
- Public modeling guides, grounded in real industrial cases
- Transparent pricing, with the first 1,000 optimized stops free to cover the integration phase
Billing aligned with what is actually operated
Kardinal charges per optimized stop, not per vehicle or per user. A stop is billed only once, regardless of how many optimizations and re-optimizations it goes through: a team or an agent can recalculate a plan as many times as needed within a 24-hour window, at no extra cost. Pricing is public and among the most competitive on the market, starting at €0.012 per stop.
***
About Kardinal:
Since 2015, Kardinal has been developing route optimization technologies for the transportation and field service sectors, combining operations research and machine learning to model the most complex business constraints. After supporting the market’s largest players through custom deployments, Kardinal is now making its expertise available through a self-service API, designed for technical teams and AI agents alike.
Kardinal was recognized as a Representative Vendor in the Gartner Market Guide for Vehicle Routing and Scheduling 2025, a recognition renewed in 2026.
For more information: kardinal.ai
Press contact:
Perrine Texier, Marketing Manager at Kardinal: perrine.texier@kardinal.ai
Note to editors:
*Data on the share of agent traffic in technical documentation: Mintlify, mintlify.com/data, overall figure across all sectors, February to August 2026.
**The rates cited are based on an internal evaluation protocol applied to modeling cases drawn from industrial projects led by Kardinal over the past ten years. Details of the protocol, the cases tested and the validation criteria are available on request from the press contact.
