AI agents that plan, route and run field operations
Cooriroo is the AI Operations Platform for fleets, distribution and field work, built in Dubai. Its agents plan the routes, watch the fleet and keep the roster. With Agent Studio you build your own, on the open A2A standard.
Swipe the map sideways to see all three lines. The platform is the station in the middle.
Everything here runs on Cooriroo, the platform
This site explains the AI. The platform is where you use it.
- Live today
Cooriroo, the platform
Fleets, distribution and field work in one place: trips, drivers, vehicles and shifts, and the agents that run them.
Open cooriroo.com
How to read this site
cooriroo.com is the product. This site explains the AI underneath it, for the people who have to trust it: operations leaders, CTOs and integrators.
Each page covers one capability: what the agent does, what it is allowed to change, and what a person still decides.
The agents
Agents that plan routes, optimise them, watch the fleet and keep the roster. They work through the same tools as the screens, under the user's own permissions.
The open standards
Agent Studio publishes every agent over Agent2Agent, so your agents and your partners' agents can find Cooriroo's and hand them work.
The city graph
Dubai2Graph joins Dubai's roads, transit, buildings and places with an operator's own trips into one structure a model can learn from.
Agent Studio: build AI agents on the open A2A standard
Agent2Agent, or A2A, is the open protocol that lets AI agents from different companies find each other and hand each other work. Google introduced it in April 2025 and donated it to the Linux Foundation, which governs it today as a vendor-neutral standard. Agent Studio is where Cooriroo's agents are built and published on it.
How an agent is built in Studio
Describe the job
Write what the agent should do in plain words, or start from a template. No code.
Set what it knows, what it may do and what needs approval
Choose its tools from the platform's catalogue. An agent can never do more than the person using it could do in the app.
Publish it with a signed Agent Card
Any A2A agent can now discover it and delegate to it: yours, a partner's, or one built with another vendor's toolkit.
What runs today. The five agents in the table are live and tested against real fleet data. The no-code builder is in development.
GET /agents/planner/.well-known/agent-card.json { "protocolVersion": "1.0", "name": "Cooriroo Planner Agent", "description": "Plans fleet operations. Recommends only; it never changes fleet data.", "capabilities": { "streaming": true }, "skills": [ { "id": "planning", "name": "Fleet planning", "examples": [ "How many drivers should we roster for next Tuesday?" ] } ] }
The card an outside agent reads before it delegates to the Planner. Shortened.
| Agent | What it does | Changes data |
|---|---|---|
| Orchestrator | Takes a request, splits it, delegates to the right specialists and returns one answer | No |
| Planner | Capacity, routing and scheduling advice, delay-risk and demand estimates | No |
| Data | KPIs, costs, fleet health, work orders, licence and insurance expiries | No |
| Logistics | Creates trips and stops, calculates routes, assigns drivers and vehicles | Yes, the only one |
| Web Intelligence | Open-web research with sources, weather and geocoding. No access to fleet data | No |
A2A connects agents to agents
The Orchestrator can hand part of a job to a partner's agent, and a partner's agent can hand a job to Cooriroo. Neither sees the other's tools or data.
MCP connects agents to tools
Every Cooriroo agent acts only through the platform's MCP server, with more than 150 tools, as the calling user. Claude or any MCP client can use the same server.
Four operations, run by agents
The agents use the same tools as the screens. Anything you can click, you can ask for, and every step is logged.
-
Route planner
Planner and Logistics agents
Pool the day's deliveries and the agent proposes how to split them across vehicles. Each run gets a driver and a vehicle and becomes a trip. Before anything is assigned, it checks for clashes with other trips, maintenance windows and driving hours.
You askCreate tomorrow's run from the Al Quoz pool with two vans.
-
Route optimization
Logistics agent
Traffic-aware stop order and arrival times on the real road network. You see the optimised route beside the current one, with the saving, before you apply it. When a stop is added or dropped mid-trip, the route is recalculated.
You askReorder trip 4821 so it finishes at the hub.
-
Fleet management
Data agent
A health score from 0 to 100 for every vehicle, built from five weighted factors. Six alert rules watch the fleet and close themselves when the problem is fixed. The live map needs no tracking box: the driver's phone is the sensor.
You askWhich vehicles must not go out tomorrow?
-
Workforce management
Planner and Data agents
Shifts planned from templates, with clock-in, breaks and overtime from the app the driver already uses. Driving-hour limits are checked live, so a near miss is flagged before it becomes a violation, and a no-show at 06:00 is an alert at 06:01.
You askHow many drivers should we roster for next Tuesday?
Dubai2Graph: the city as one graph
Roads, bus stops, buildings, businesses and a distributor's day of deliveries, assembled into one graph of Dubai. It is the map the routing agents will learn from. Nine figures, all from open data.
What the graph already shows
Three of the six results. Each is stated with its limits on the Dubai2Graph page.
- 1,614 stops against 279
- Reachable by public transport within an hour from Terminal 3 at DXB, and from Jebel Ali Free Zone. The airport is served six times better than the largest logistics zone.
- Three logistics bodies, not one
- Deira and the Creek, Al Quoz, and the Jebel Ali corridor, with near-empty space between them. Territory design starts there, not from a grid.
- 89 km to 427 km per van
- One synthetic distributor's day, 120 real addresses, same load per van. The plan shows the long sector at 06:00 instead of 16:00.
Built with City2Graph, the open-source library for turning cities into graphs, and PyTorch Geometric.
Pages on this site
One page per capability. Each one opens from here.
- Flagship
Agent Studio
The five agents, the A2A standard, the Agent Card, and how an agent is built, tested and given more autonomy.
- Capability
Route planner
From a pool of orders to assigned trips, with the checks the agent runs before it proposes a plan.
- Capability
Route optimization
How stop order and arrival times are computed, previewed and recalculated during the day.
- Capability
Fleet management
The vehicle health score, the six alert rules and the live map without hardware.
- Capability
Workforce management
Rosters, driving-hour limits and what happens in the minute after a no-show.
- Blog
Mobility and AI
Short pieces on how cities move and how agents do operational work.
- Research
Dubai2Graph
Dubai as one graph: nine figures, six results, the method and the data behind each.
Blog: mobility and AI
Short pieces on how cities move and how agents do operational work, written from our own data.
- Mobility
Why Jebel Ali is six times harder to reach by bus than the airport
One hour of public transport from Terminal 3 reaches 1,614 stops. The same hour from Jebel Ali Free Zone reaches 279.
See it run on your operation
Forty-five minutes with an engineer. Bring one real routing or dispatch problem and we work through it with the agents.
Your request goes to our team directly. No newsletter. See the privacy note.
Request sent
An engineer will reply within one working day with two time slots and a question about the problem you want to see solved.