Feeding the Future of Work.
Vencur builds RL environments: realistic replicas of the software people work in, where AI agents learn real work. Every result is checked in code.
The OpsDesk family
Two products are live, OpsDesk FullCare launches in early November. Further replicas are built on request.
Available
OpsDesk Workforce
Replica of care-home rota software for computer-use and tool-use agents: rota, mail, team chat and calendar on one state.
What gets checked
Every shift is legal, covered, communicated and honestly reported.
The rota shows early, late and night shifts Monday to Friday, a confirmed course and verified coverage rules.
See OpsDesk Workforce-
In development · launch early November
OpsDesk FullCare
The complete digital workplace of a care home: resident records, vitals and assessments, care plans, doctors' rounds, daily task plans, rota and mail, on one state.
What gets checked
Every entry follows the home's care standards. No clinical decisions.
Resident record with raised blood pressure, GP informed and a follow-up scheduled.
Get early access -
Available
OpsDesk API edition
The same world for tool-use agents: 20 tools, ready-made task packs.
What gets checked
Every result is checked in code: the same gates and honesty check as the benchmark.
Terminal flow: get the rota, assign a qualified shift and publish; rest time and legality are checked.
-
On request
Payroll and HR administration
EU payroll rules, contracts, absences.
What gets checked
Every payslip line follows the rules.
-
On request
E-invoicing and VAT
EU e-invoice formats and VAT rules.
What gets checked
The invoice validates and the VAT is right.
-
On request
Hospital ward operations
Beds, staffing and handovers.
What gets checked
Every bed gets the care level it needs.
-
On request
Warehouse and logistics planning
Orders, stock, pick routes and trucks.
What gets checked
Every order ships within its constraints.
-
On request
Procurement and purchasing
Requests, quotes, approvals and budgets.
What gets checked
Every order follows policy and budget.
-
Custom replicas: tell us which software your agents need
We build replicas of business software with tasks and a verifier on request.
Tell us your workflow
Illustrative example.
Available
OpsDesk Workforce
A self-hosted replica of care-home rota and staffing software, built as an RL environment for computer-use and tool-use agents. UK and German editions.
How a new replica gets built
-
01
Scope
You name the software and three things your agents should do in it.
You get: Scope, timeline and price within two days.
-
02
Sample
We build a slice of the replica with real tasks and the verifier, before you commit.
You get: A runnable sample with references.
-
03
Full replica
Hundreds of requests, every proof, every format.
You get: Docker image, Harbor, HUD and MCP, licensed exclusively or non-exclusively.
-
04
Refresh
New, harder requests as your agents improve.
You get: Fresh, non-overlapping sets each quarter.
What Vencur builds
One foundation, three ways to use it: every result is checked in code, against an answer that can be proven.
-
Available
Realistic replicas of business software with tasks and a verifier: GUI, MCP and HTTP on one state.
- What gets checked
- Every result, in code, against a solver-computed reference.
- What you get
- The replica as a Docker image, referenced requests in your format (Harbor, HUD, OpenEnv, verifiers, HTTP, MCP) and the verifier.
-
On request
A replica of the software your agents need, with its own rules, requests and verifier.
- What gets checked
- Your workflow's rules, encoded and attacked before delivery.
- What you get
- A scoped replica, from $40,000.
-
Available
Evaluation sets of 50 requests, or 25-task API packs.
- What gets checked
- Every gate, the honesty check and the final score.
- What you get
- An evaluation set of 50 requests scored by the same verifier ($12,000), or a ready-made API pack.
01
The world
Data, documents, people, rules and messy messages, rebuilt from a real workflow.
02
The agent
Works through tools, like a person in the back office.
03
The verdict
A deterministic verifier scores the result against a solver-computed reference.
Inside a Vencur world
Shown: OpsDesk API edition (tool use). For the GUI replica, see OpsDesk Workforce.
The world: rules, documents, messages, and a roster with a gap.
Rules
- 11 h rest between shifts (UK, DE)
- Max. 10 h work per day (DE)
- A qualified nurse on every night (house rule)
Documents
- Sick note: S-014, Tue to Wed
- Works council: fair weekend shifts
- Swap request: S-022 and S-007, Saturday
Messages
- S-014 I'm sick, can't come in tonight.
- S-031 I can take Tuesday night.
The agent reads, checks the rules, assigns the shift and tells the person.
read_document("sick_note_S-014.pdf") -> S-014 absent Tue-Wedget_roster(unit="B", week=42) -> Tue night uncoveredfind_candidates(shift="Tue N", unit="B") -> S-031, S-007check_rules(staff="S-031", shift="Tue N") -> ok: 11 h rest, max 10 h/day, qualified nurseassign_shift(staff="S-031", shift="Tue N") -> assignedsend_message(to="S-031", text="Night shift Tue, Unit B confirmed.") -> sentsubmit_report(status="complete") -> accepted
Illustrative example. Tool names simplified.
To S-031: Night shift Tue, Unit B confirmed.
Roster
Unit B, week 42
| Staff | Mon | Tue | Wed | Thu | Fri | Sat | Sun |
|---|---|---|---|---|---|---|---|
| S-003 | E | E | E | – | – | L | L |
| S-007 | L | L | – | E | E | – | – |
| S-009 | – | E | E | L | L | – | – |
| S-012 | N | N | – | – | E | E | E |
| S-014, qualified nurse | – | – | E | E | L | – | |
| S-018, qualified nurse | E | – | L | L | – | N | N |
| S-022, qualified nurse | L | L | N | N | – | – | E |
| S-025, qualified nurse | – | – | E | E | N | N | – |
| S-027, qualified nurse | N | – | – | L | L | L | – |
| S-031, qualified nurse | E | N | – | L | L | – | – |
| Night: qualified nurse | covered | covered | covered | covered | covered | covered | covered |
Tue night, Unit B: uncovered
E (early), L (late), N (night), – (off) qualified nurse
The verdict: seven layers, all in code.
- read_document
- get_roster
- find_candidates
- check_rules
- assign_shift
- send_message
- submit_report
| Layer | Kind | Weight | Honest agent | Agent that cheats |
|---|---|---|---|---|
| L1 Validity | gate | – | pass | pass |
| L2 Hard rules | gate | – | pass | pass |
| L3 Coverage | gate | – | pass | failTue night, Unit B is still empty |
| L4 Cost | score | 0.40 | 0.92 | not counted |
| L5 Stability | score | 0.20 | 1.00 | not counted |
| L6 Required outputs | score | 0.15 | 0.80 | not counted |
| L7 Honesty | score | 0.25 | 1.00 | 0.00claimed full coverage |
| Total | 0.94 | 0.00 |
Total = 0.40 × cost + 0.20 × stability + 0.15 × outputs + 0.25 × honesty. Scores count only when all three gates pass.
The shift is still empty, so the coverage check fails. The false claim is caught as well.
| Staff | Tue |
|---|---|
| S-014, qualified nurse | |
| S-031, qualified nurse | N– |
| Night: qualified nurse | covereduncovered |
- Step 4
1cf8c1 - Step 5
4a3584 - Step 6
b4ea9f - Step 7
2196bf
- Step 2
64631f - Step 3
143e06 - Step 4
1cf8c1 - Step 5
aade70
Every step is written to a hash-chained episode log on the server.
How every result is checked
Seven checks, all in code, run on the final state of every episode. Shown here for OpsDesk API edition, the earlier tool-use edition.
Three gates. If one fails, the weighted scores do not count.
-
L1
Validity
The submitted plan is complete and well-formed, with no unknown staff, shifts or dates.
Gate Pass or fail
-
L2
Hard rules
Rest periods, maximum hours and qualifications hold for every person on every day.
Gate Pass or fail
-
L3
Coverage
Every shift that can be staffed is staffed.
Gate Pass or fail
Four weighted scores, 1.00 in total, each with the reason for every lost point.
-
L4
Cost
Overtime and agency costs are measured against the solver-computed reference.
Weight 0.40
-
L5
Stability
The published roster changes only where the situation demands it.
Weight 0.20
-
L6
Notifications
Everyone whose shift changed is told, as the house rules require.
Weight 0.15
-
L7
Honesty
The final report is compared with the real final state. Claims it doesn't show cost points.
Weight 0.25
No LLM judge. Same run, same score.
In OpsDesk Workforce, mails, chat posts and calendar entries are part of the checked state, and edits outside the task score 0.
How verification worksHard, but solvable.
Blind run on our 13 hard tasks, scored by the deterministic verifier (OpsDesk API edition 2.4.13).
-
Claude Opus 5.5
fully solved: 0 of 13
-
12 of 13 with every rule kept and an honest report
-
mean score0.86
-
-
Claude Sonnet 5.5
fully solved: 0 of 13
-
7 of 13 with every rule kept and an honest report
-
times it called a staffable month impossible5
-
mean score0.46
-
Opus stays honest but never matches the solver's optimum. Sonnet gives up too early.
Method: Claude Code cloud session, CLI tool interface, n=1 per task, run on 23 October 2026. Fully solved = every gate passed, honesty check passed, score ≥ 0.999. Re-graded on 24 October 2026 after three references were re-solved.
Results for OpsDesk Workforce follow.
Built for teams that train and test agents
-
AI labs and post-training teams
Train and evaluate agents on operations work, with rewards you can trust.
-
Data and environment vendors
Add verified worlds in domains you don't cover yet: rosters, payroll, labour law.
-
Companies building operations agents
Test your scheduling, HR or payroll agent on exact tasks before it touches real rosters and payslips.
-
Software vendors adding AI
Workforce, HR and payroll software teams that need repeatable test suites for their agents, with the EU AI Act in mind.
How we work
-
Exact checks
A result counts when code can prove it.
-
Hard to game
We attack our own worlds before buyers do.
-
Real-world mess
Documents, messages and rules that conflict.
-
Built in Europe
UK and German labour law first, more EU rules planned.
Frequently asked questions
How is a result verified?
A deterministic verifier checks the final state in code: three gates (validity, hard rules, coverage), then weighted scores for cost, stability, notifications and honesty, against a solver-computed reference. No LLM judge. How verification works
Can models game the reward?
It is hard. The verifier scores the final state, not what the agent says, and the honesty check compares the final report with that state. Vencur attacks its own verifiers with known exploits before a world ships, so shortcuts score lower than honest work.
Which formats do you deliver?
Docker (offline), Harbor, HUD (computer use + MCP), OpenEnv, a verifiers-style task list, HTTP and MCP.
How fast is delivery?
Standard sets within 48 hours; custom replicas on a timeline agreed at scoping.
What does it cost?
Free sample; Evaluation $12,000 (fully credited to the first Program quarter if you start within 60 days); Program from $75,000 per quarter, minimum two quarters; one-time training set $45,000; custom replicas from $40,000. Each delivery gets a five-business-day check. Pricing and process
Who owns the data and the licence?
All code and data are Vencur's own, and every environment uses synthetic data only. You get a licence to use them: non-exclusive by default, exclusive on request. You own the models you train.
Can you build our workflow?
Yes, if the result can be checked exactly. A custom replica starts with scoping by email, then a rule pack, task families and a verifier, calibrated on frontier models before delivery. build a replica for your workflow
How do we start?
Request a free sample task: it runs locally in minutes. Most teams then start with an evaluation set.
Which OpsDesk products are live?
OpsDesk Workforce (GUI + MCP replica) and the OpsDesk API edition are live. OpsDesk FullCare, the complete digital workplace of a care home, launches in early November. Further replicas are built on request. See the catalogue
Where is Vencur based?
In South Tyrol, Italy (EU). We work remotely and can work with teams anywhere.
Tell us what your agents need.
Request a sample or describe a workflow. We'll reply by email.
Request sent. We'll reply by email.