JustRouting MCP Server
io.github.justroutingv0.1.5Updated Oct 3, 2026
Road routing, travel-time matrices, geocoding, and fleet optimization for Southeast Asia.
Overview
Lets an assistant geocode places, calculate driving or motorcycle routes, build travel-time matrices, and optimize multi-vehicle delivery plans in Southeast…
- What it does
- Wraps the JustRouting routing API in four tools: geocode converts place names or addresses into coordinates; route returns driving distance and duration between two points; table builds a distance and duration matrix across many locations; optimize assigns jobs to vehicles and orders each vehicle's stops. Driving is the default profile, with a motorcycle profile available. The server is a thin stdio process that forwards requests to the JustRouting API.
- When to use it
- Use it when an assistant needs road routing, travel times, or fleet planning in Southeast Asia, such as comparing which driver is nearest to a customer, planning multi-van pickup routes, or answering how long a drive between two places takes. It is not useful outside the covered region.
- Requirements
- A local process started by the MCP client over stdio. A JustRouting API key is required and passed in the JUSTROUTING_API_KEY environment variable. Prebuilt binaries exist for macOS, Linux, and Windows; building from source needs Go 1.25 or later. Network access to the JustRouting API is required.
Installation
In SourceWeft
- Open JustRouting MCP Server in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
JustRouting MCP
MCP server for JustRouting, a Southeast Asia-focused routing API.
Use JustRouting's road-routing capabilities from MCP-compatible AI assistants such as Claude and Cursor.
Features
- Driving route calculation
- Motorcycle routing profile
- Driving distance
- Estimated travel duration
- Distance and duration matrix (table)
- Address and place search (geocoding)
- Route optimization (vehicle routing)
- Southeast Asia-focused road coverage
- MCP stdio transport
- Built on the official JustRouting Go client
Requirements
- A JustRouting API key
- Go 1.25+ (only needed when installing from source)
Installation
One-line installer (macOS / Linux)
Downloads the latest prebuilt binary for your OS/architecture from GitHub Releases, verifies its checksum, and installs it to /usr/local/bin (or ~/.local/bin). No Go required.
Pin a specific version:
Uninstall: remove the justrouting-mcp binary from the directory it was installed to.
Manual download
Prebuilt binaries for macOS, Linux, and Windows are attached to every release. Download the archive for your platform, extract it, and put the binary on your PATH.
From source (Go users)
Make sure justrouting-mcp is available in your PATH.
Configuration
Set your JustRouting API key:
Tools
route
Calculate a route between two locations, using the driving profile by default or the motorcycle profile when requested.
Input:
profile is optional. Set it to "motorcycle" for a motorcycle route; omit it (or use "car") for the default driving route. When the user mentions a motorcycle or motorbike, the assistant sets profile to "motorcycle".
Coordinates must use:
If the user asks about places by name or address instead of coordinates, call geocode first (see below) and pass its coordinates values here.
Output:
distance_meters is the driving distance in meters.
duration_seconds is the estimated driving duration in seconds.
geocode
Search for places and convert a place name or address into coordinates. Use this before route when the user refers to places by name.
The search is structured: the assistant parses the user's place reference into address components and passes only the ones it can determine — name, housenumber, street, postcode, city, country. Components with no data are omitted. Prefer including country (and city) when the context implies them — they are the strongest disambiguators for common, abbreviated, or misspelled names.
Input:
At least one component is required. limit is optional and defaults to 1 (best match only); it must not exceed 10. filters is also optional, for example "filters": ["countrycode:sg"] to restrict results to Singapore.
Output:
Results are ordered best first — always take the first result's coordinates (ready to pass to route, table or optimize). If nothing matches, the tool returns an error.
table
Calculate a matrix of driving durations and distances between many locations at once. Use it to compare several places — for example, to find which of several drivers is nearest to a customer.
Input:
coordinates is required and must hold at least two entries in longitude,latitude format (as returned by geocode). The position of each entry is its index: matrix rows and columns, and the index field of every source/destination in the output, refer back to it. Pass the places in a fixed order — for a nearest-driver question, put the customer first and the drivers after, then read the first row.
sources and destinations optionally restrict the matrix to subsets of the coordinates by index (empty or omitted means all of them). annotations optionally selects "duration", "distance", or both (default both). profile works exactly as in route — set "motorcycle" when the user mentions a motorcycle or motorbike, otherwise omit it (or use "car") for driving.
Output:
durations is in seconds and distances in meters, each indexed [source][destination]. A null entry means the engine could not connect that pair — it is not zero. Every source and destination object carries an index back to the input coordinates list.
optimize
Assign jobs to vehicles and order each vehicle's stops (vehicle routing). Use it for delivery or visit planning — for example, two vans picking up parcels from customers.
The assistant geocodes every place first — always taking each call's first (best) result — then builds the request. For a prompt such as "I have two vans, V1 at 'garlick ville singapore' and V2 at 'victoria place singapore'; pick up parcels from 6 customers at 'original sin', 'henry park primary school', 'astrid meadows tennis court', 'little oaks montessori kindergarten', 'eden hall' and 'villa chancery'", the assistant geocodes all 8 places and calls:
Input:
At least one vehicle and one job are required, with unique ids. Coordinates are longitude,latitude strings (as returned by geocode). Set a vehicle's start and end both to its current location for a round trip, or omit them (or pass "") when the vehicle may start or end anywhere. profile works per vehicle exactly as in route — set "motorcycle" when the user mentions a motorcycle or motorbike, otherwise omit it (or use "car") for driving.
Output (vehicle 2's route and some fields omitted for brevity):
Read each route's steps in order: "job" steps carry the job id plus the arrival time and travel duration in seconds, so they tell which vehicle serves which jobs and when. unassigned lists the jobs no vehicle could serve. summary aggregates cost, duration and distance across all routes.
Asking about places by name
For a prompt such as "how long from 'marina bay singapore' driving to 'changqi airport'?", the assistant geocodes each place and then routes:
geocodewith"name": "marina bay", "country": "singapore"→ take the first result'scoordinatesgeocodewith"name": "changqi airport"→ take the first result'scoordinatesroutewith the twocoordinatesvalues asoriginanddestination(omitprofilefor driving)
For a comparison such as "which of these 4 drivers is nearest to the customer?", geocode the customer and every driver, take each call's first result, pass all coordinates values to table in a fixed order — customer first — and read the first row of the returned matrices.
For a multi-vehicle plan such as "two vans, six customers to pick up from", geocode the vans' locations and every customer, take each call's first result, then call optimize with one vehicle entry per van (start and end set to its location) and one job entry per customer. Read the routes[].steps in order to see which van serves which customers.
Claude
Configure the MCP server:
Cursor
Add the following MCP server configuration:
Development
Clone the repository:
Install dependencies:
Run tests:
Build:
Run:
The server communicates with MCP clients through stdin/stdout.
Releasing
Tag and push — GitHub Actions builds the binaries and publishes the release:
The install.sh script always installs the latest release, so it never needs to be updated when a new version ships.
Architecture
The MCP server is intentionally thin.
Routing logic, API authentication, HTTP transport, retries, and API error handling are provided by the official JustRouting Go client.
Roadmap
- Route geometry
- Alternative routes
- Waypoints
- Distance matrix
- Route optimization
- Streamable HTTP
- Remote MCP deployment
License
Source: README.md at commit 5ac78d3
Tools
0Version history
1- v0.1.5LatestOct 3, 2026
More servers like JustRouting MCP Server
CvLAC (MinCiencias)
io.github.stivenson
Lets an AI assistant read and update a Colombian CvLAC (MinCiencias) research CV by driving the CvLAC website with your own account.
Appmetrica Mcp
io.github.pavellunev
Lets an assistant read Yandex AppMetrica mobile analytics — reports, funnels, segments, dashboards, raw logs and push stats — and optionally create funnels…
Noeto
io.github.noeto-tasks
Lets an assistant read and work a Noeto kanban board: browse boards, search cards, create, update, move and comment on cards, and keep documents on them.