Set up your GCC. Run it on a platform built for one.
One offering, end to end: strategy, entity and location, hiring, onboarding and daily operations. ScriptBees has set up and run capability centres since 2013, and GCC-in-a-Box is the platform they run on.
- POST /v1/agents/hr-agent/runs · actor: gcc-manager
- Identity resolved · token ttl 10m · org: northwind
- OPA policy evaluated → REQUIRE_APPROVAL (production-guardrails)
- Approval granted by admin@northwind · single-use · sod: ok
- Gateway call executed · egress allow-listed · 212ms
- Audit record appended · chain verified [OK]
GCC setup, end to end
One accountable team from the first question to a running centre. Each stage has a product behind it.
- PlanLocation, functions, headcount, budget and state incentives become a graded strategy.
- LaunchEntity, facilities and a second person's approval. One transaction builds the centre.
- HireSeats proposed within capacity. Screening in bands. Offers wait for a person.
- OperateOnboarding, business agents and eight integrations carry the daily load.
- ScalePerformance, engagement and a rostered AI workforce with named owners.
Every incentive you qualify for. Claimed, not just listed.
Nine states run dedicated GCC policies. We map your eligibility, shape the plan to qualify for the most, and file and track every claim to disbursement.
Up to 20–30% of capex and 15–20% of operating spend, with caps that rise with the size category you land in.
Three-year reimbursements on high-salary roles, sequenced so senior hires fall in the richest year.
Stamp duty, electricity duty, SGST and training reimbursements, each filed inside its window.
Two products, built for GCCs
The platform runs the centre. The control plane governs every agent in it.
The GCC platform
- Ten-step launch wizard, graded by fifteen checks
- Hiring with guardrails: agents score, people decide
- Seven business agents for HR, IT, finance and more
- A Command Center where no figure is invented
Governance under your GCC
- Identity, policy and approvals for every agent in the centre
- Every action resolves to allow, deny or ask a human
- Append-only, hash-chained audit HQ can verify
- Cost budgets that warn before they refuse
Built for the questions an auditor will ask
People decide
Agents propose. No agent hires, rates or approves.
Evidence, not trust
A hash-chained audit trail your auditors can verify themselves.
No implicit access
Every tool, peer and memory needs an explicit grant.
Fails closed
An unreachable policy engine is a deny, never an allow.
No fake numbers
A figure that cannot be computed shows "—", never zero.
Open standards
MCP, A2A, OIDC, OPA and OpenTelemetry. No lock-in.
Find your starting point in thirty seconds
Three questions, one recommendation.
Governance below, operations above
GCC-in-a-Box
Launch, hiring, lifecycle, operations.
HQ systems
Slack, Teams, Jira, ServiceNow and more, through held integrations.
Your centre's own agents
Built by your teams, governed the same way.
Agent Control Plane
Identity → Policy → Budget → Approval → Execution → Audit → Cost
Any model
Adapters name no provider.
Any OIDC provider
Single sign-on for people, short-lived tokens for agents.
Your cloud, your region
Kubernetes, Vault, data residency built in.
Straight answers
Can an agent hire, reject or rate someone?
No. Agents score and draft; advancing, rejecting, offering and rating are held for a person with the permission. There is no flag that changes this.
Do we have to replace the AI tools we use?
No. The control plane is not a framework, a model or a runtime. It governs the agents you already run.
Where does our data live?
In your cloud, in your region. Data residency can be set to off, advisory or strict at every point data could leave.
What will HQ and our auditors see?
A Command Center computed over your data, and an append-only, hash-chained audit log verifiable through the API.
Does it work with our HR and IT systems?
Slack, Teams, Microsoft 365, Google Workspace, Jira, GitHub, Zoom and ServiceNow ship today, with no credentials stored. Applicant tracking has a seam for Greenhouse or Workday.
What is honestly not there yet?
Account provisioning records decisions rather than creating accounts, search is lexical rather than semantic, and agent cost is modelled, not metered. We will show you the full list.
Start with a forty-five minute setup call
Bring a location, a headcount and a budget. We draft your centre's strategy live and map the setup plan together.