For ERP/MRP vendors and IT integrators
Commercial access to the high-performance calculation engine for ERP/MRP developers and IT integrators. REST in, optimized sequence out.
Your ERP already knows the routings, the machines and the due dates. It just cannot schedule them. Post that data to the commercial solver gateway, poll the job and receive a feasible, optimized sequence. Two input formats: structured JSON or compact DZN.
Every call carries a B2B bearer token: Authorization: Bearer <your_api_key>. Solving is asynchronous — POST returns a job_id, GET returns the status and the result.
Solver input is passed as a structured model_data object instead of raw DZN text. The gateway includes a normalization layer, so you write plain developer notation and the engine converts it to the set structures the MiniZinc compiler requires.
/api/commercial/v1/solve/json— Start solving (asynchronous)/api/commercial/v1/solve/json/{job_id}— Job status and results3 machines, 2 jobs, 4 tasks, 6 operations — submitted with one extra constraint and a 60 s time limit.
Request
curl -i -X POST https://api.makespan.pl:443/api/commercial/v1/solve/json \
-H "Authorization: Bearer $MAKESPAN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"schedule_id": "json-small-schedule-uuid",
"user_id": "json-small-user-uuid",
"model_data": {
"no_mach": 3,
"no_jobs": 2,
"no_task": 4,
"no_optt": 6,
"tasks": [[1, 2], [3, 4]],
"optts": [[1, 2], [3], [4, 5], [6]],
"optt_mach": [1, 2, 2, 2, 3, 1],
"optt_dur": [10, 15, 20, 12, 8, 5]
},
"constraints_text": "constraint start[1] >= 5;",
"time_limit_seconds": 60
}'Response
202 Accepted{
"job_id": "job-0c19ea48",
"status": "queued",
"schedule_id": "json-small-schedule-uuid",
"estimated_wait_seconds": 9
}Poll the job_id returned by POST. Once status is completed, result_json carries the optimized schedule.
Request
curl -i -X GET https://api.makespan.pl:443/api/commercial/v1/solve/json/job-0c19ea48 \
-H "Authorization: Bearer $MAKESPAN_API_KEY"Response
200 OK{
"job_id": "job-0c19ea48",
"schedule_id": "json-small-schedule-uuid",
"status": "completed",
"error_message": null,
"started_at": "2026-08-22T16:43:40.947288Z",
"finished_at": "2026-08-22T16:43:41.242836Z",
"result_json": {
"start": [5, 15, 0, 15],
"dur": [10, 20, 8, 5],
"b": [true, false, true, false, true, true],
"objective": 35
}
}This format accepts solver input written as raw DZN text (where newline characters are encoded as \n).
/api/commercial/v1/solve— Start solving (asynchronous)/api/commercial/v1/solve/{job_id}— Job status and resultsThe same model — 3 machines, 2 jobs, 4 tasks, 6 operations — sent as a single dzn_text string.
Request
curl -i -X POST https://api.makespan.pl:443/api/commercial/v1/solve \
-H "Authorization: Bearer $MAKESPAN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"schedule_id": "14e6636f-8388-4e57-8dfd-f428c5fe1190",
"user_id": "79f04043-fec5-4081-b276-774412df0821",
"dzn_text": "no_mach = 3;\nno_jobs = 2;\nno_task = 4;\nno_optt = 6;\ntasks = [1..2, 3..4];\noptts = [1..2, 3..3, 4..5, 6..6];\noptt_mach = [1, 2, 2, 2, 3, 1];\noptt_dur = [10, 15, 20, 12, 8, 5];",
"constraints_text": "constraint start[1] >= 5;",
"time_limit_seconds": 60
}'Response
202 Accepted{
"job_id": "job-0c19ea48",
"status": "queued",
"schedule_id": "14e6636f-8388-4e57-8dfd-f428c5fe1190",
"estimated_wait_seconds": 9
}The finished job returns result_dzn: objective plus the start, dur and b arrays in DZN notation.
Request
curl -i -X GET https://api.makespan.pl:443/api/commercial/v1/solve/job-0c19ea48 \
-H "Authorization: Bearer $MAKESPAN_API_KEY"Response
200 OK{
"job_id": "job-0c19ea48",
"schedule_id": "14e6636f-8388-4e57-8dfd-f428c5fe1190",
"user_id": "79f04043-fec5-4081-b276-774412df0821",
"status": "completed",
"result_dzn": "objective = 35;\nstart = [5, 15, 0, 15];\ndur = [10, 20, 8, 5];\nb = [true, false, true, false, true, true];",
"error_message": null,
"started_at": "2026-08-21T14:03:49.924494Z",
"finished_at": "2026-08-21T14:03:50.435426Z"
}For ERP/MRP-class production systems, webhooks are the recommended integration method with makespan.online instead of continuously polling for job status.
Production schedule optimization is computationally heavy. Depending on problem size and the configured time_limit_seconds, a run can take from a few seconds to several minutes.
Simply configure a Webhook URL in your account settings (Tenant Portal), e.g. https://your-erp-system.pl/api/v1/makespan-callback. Our server will send a POST request there with Content-Type: application/json.
Best practice: Your receiving server should respond immediately with status 200 OK or 202 Accepted (ideally within 2 seconds). Process and store the result asynchronously in the background to prevent the gateway from closing the connection due to timeout.
When the original job was sent in JSON format, the webhook returns the result in result_json.
{
"job_id": "job-ccc5892d",
"schedule_id": "json-small-schedule-uuid-vps-test",
"status": "Success",
"error_message": null,
"solver_seconds": 12.35,
"finished_at": "2026-08-26T06:59:44.125Z",
"result_json": {
"start": [5, 15, 0, 15],
"dur": [10, 20, 8, 5],
"b": [true, false, true, false, true, true],
"objective": 35
}
}When the original job was sent in DZN format, the webhook returns the result in result_dzn.
{
"job_id": "job-0c19ea48",
"schedule_id": "14e6636f-8388-4e57-8dfd-f428c5fe1190",
"status": "Success",
"error_message": null,
"solver_seconds": 18.12,
"finished_at": "2026-08-26T07:11:05.435Z",
"result_dzn": "objective = 35;\nstart = [5, 15, 0, 15];\ndur = [10, 20, 8, 5];\nb = [true, false, true, false, true, true];"
}Two endpoints are all you integrate: POST to start solving, GET to collect the status and the result. No agents, no on-prem installation, no schema migration on your side.
Solving is asynchronous, so a long run never blocks your ERP transaction.
Webhooks remove the need to poll GET /solve/{job_id} every few seconds. Your ERP sends one POST and resumes work only when the gateway pushes the completed result.
The gateway pushes a notification as soon as the solver finishes — whether the job succeeds, times out or fails validation — so your system reacts to real events instead of checking for them.
Background integration without blocking worker threads. Your application scales better because it does not hold connections open while the engine is optimizing.
As soon as the solver finds the first schedule satisfying every constraint, it does not stop. It keeps searching for a better objective — for example a shorter makespan — until the granted time_limit_seconds is exhausted.
You always receive the best schedule found within your time budget.
Counter-intuitively, adding constraints usually shortens the solve, not lengthens it.
Technically: every constraint prunes the search tree of the optimization problem, so the remaining space to explore is smaller and the search moves through it faster.
A free sandbox key, deterministic sample datasets and complete developer documentation (Swagger UI and ReDoc) for your integration team.
The solver constraint library currently counts around 10 items. Next to simple ones — such as "job A may start between date X and date Y" — it covers several advanced constraints:
A flat developer base fee plus solve-volume tiers, so you can price your own APS add-on with confidence.
Request free beta access and bring one week of real production data.