Launching today

Kardinal

Launching today

Route optimization API built for developers and their agents

14 followers

Route optimization API built for developers and their agents

14 followers

Visit website

Kardinal is a route optimization API rebuilt for AI agents to call directly: typed SDKs, docs an agent can actually read. Built on 10 years of production route optimization, for the constraints that break most "demo" solvers (e.g. capacity-gated returns, self-closing consolidation loops, key-holder-only access windows). Test it entirely on your own: free tier, no sales call, no quote. The most complete API on the market: 93% of constraints covered, per our 2026 benchmark of 11 APIs.

Free Options

Launch tags:API•Developer Tools•Artificial Intelligence

Launch Team / Built With

Maker

📌

Hey Product Hunt community 👋 I'm Cédric, co-founder of Kardinal. For the last 10 years, we've solved route optimization for companies like DPD, GLS, and JCDecaux, as a service: they'd send us their data, we'd configure a solver, they'd get routes back. It worked well, but it meant months of sales calls before anyone could tell if it actually fit their case. Somewhere in the last year, something changed in the requests we got. More and more weren't coming from ops teams anymore, they were coming from developers trying to wire an AI agent into their logistics stack. That exposed a problem we hadn't fully seen before. An agent working from documentation written for humans produces code that compiles, API calls that succeed, and a model that's wrong. Not wrong in a way that throws an error. Wrong in a way that looks fine on a map and falls apart on the road. A plan can look perfectly optimized and still be physically impossible to run, because a large customer return (a sofa, a bed) picked up early in the route ends up blocking the vehicle for every delivery after it. Nothing in the API response flags this. The request was valid. The model just didn't capture the constraint that mattered. So today we're opening Kardinal as a self-serve API, rebuilt specifically for that failure mode. WHAT IS KARDINAL? A REST API that takes your stops and vehicles and gives you optimized routes back in under 2 seconds. We cover 139 of the 149 real-world constraints in our own benchmark (93%), pulled from actual client cases: time windows, multi-depot, driver breaks under EU 561, cold chain, pickup-delivery pairing, heterogeneous fleets. WHY IS IT DIFFERENT? Most routing APIs are built for human developers to read. We rebuilt ours for agents using what we internally call harness engineering: multiplying the contact points between an agent and the right documentation, so it gets the right information exactly when it needs it. Concretely: an API doc dedicated to agents, a CLI, a typed SDK that makes certain mistakes impossible to write, and docs that explain when to use a parameter, not just what it does. We tested this internally before shipping it. Our own client optimization studies are now largely run by agents working alongside our solution engineers. On our internal benchmark, that took agent modeling failure rate from over 80% down to under 5%. WHO IS IT FOR? Developers and ops teams building routing into their own product or workflow, whether that's a human-built integration or an agent-driven one. WHAT CAN YOU DO TODAY? - Create an account and get an API key, no sales call, no quote - Send stops and vehicles, get optimized routes back in under 2 seconds - Re-optimize from live state in under a second if a truck breaks down mid-route, free, unlimited within 24h - 1,000 free stops a month to test your actual case, at some of the most competitive pricing on the market Billing is per optimized stop, not per vehicle or per user, so a team or an agent can recompute a plan as many times as it needs without worrying about the cost of iterating. Would love your feedback, especially if you've tried wiring routing into an agent workflow and hit a wall, or fought with a routing API on constraints it didn't fully support. I'll be in the comments all day.

Report

3d ago

Maker

Adding some technical color since I built the core engine: the hard part wasn't the routing algorithm, it was making 149 different real-world constraints composable without the request payload turning into a mess. That's what the typed SDKs are actually for, not just convenience, they catch invalid constraint combinations before the request even goes out. Happy to go deeper on any of it here.

Report

6h ago

Kardinal is riding the wave and adapting to today's changing world. As AI is growing and layers are reduced, accessing Kardinal's optimization engines has never been easier or more open to everyone. All people in the field must give it a try.

Report

6h ago

Maker

Something we're watching closely: as LLM usage keeps growing, we're seeing fewer teams look for one platform that does everything, and more teams composing their own stack from specialized pieces, each one really good at a narrow job. Routing is a good example of why: it's a hard enough problem on its own that bolting it onto a general-purpose tool rarely holds up once real constraints show up. We built Kardinal to be one of those pieces, not a platform to lock into. Curious whether others here are seeing the same shift toward composable stacks versus all-in-one tools.

Report

6h ago