The Center of Agentic platform

# Build, test and operate enterprise agents

From the first business outcome to the next governed release. One platform to create useful agents, check their authority and manage the work they do.

01 / Build

## Business context becomes a working agent.

Describe an outcome, start with a domain pack, and connect the approved knowledge and systems the work needs. Agent Builder guides you through six steps, from definition to release.

- Six-step Agent Builder

- Domain packs and knowledge sources

- Marketing Cloud and Jira connections

### Domain model

The concepts, terminology, workflows and risks shared by agents in a business area.

Campaign briefs, audience definitions and release checks.

### Business context

Company policies, approved claims, systems and data models that describe how your organization works.

Brand guidelines, product catalogs and Marketing Cloud schemas.

### User and task context

The active role, workspace, approved data access and purpose of the current task.

A coordinator prepares a campaign within their granted scope.

02 / Test

## Test what must never happen.

Check the boundaries before release. Policy simulations use the same engine as governed runs, without calling a model. Required tests and domain-pack baselines can block publishing.

- Policy What-If Simulator

- Datasets and regression experiments

- Required tests and publish gates

Authority tests ask whether an action is permitted. Output-quality tests ask a separate question: whether an allowed answer is useful.

03 / Operate

## Keep the work visible.Keep the business in control.

Run agents in Studio. Review held operations. Inspect run receipts and changes in authority across the fleet. Turn reviewed feedback into future tests.

### Run and review

Studio, Monitoring and Review Queue bring the work and its exceptions together.

### Inspect the evidence

Separate the policy verdict from the execution outcome. See available usage and accounting gaps.

### Improve with oversight

Reviewed corrections become tests. Agents do not silently change their own permissions.

Where it works

## One platform. Clear availability.

Use the right surface for the work. Available capabilities, alpha experiences and planned surfaces are identified here.

### Web workspace

Build, test, run and review agents in the browser, with tenant-scoped access and approval workflows.

### macOS desktop

Project-scoped local files and cloud-proxied model calls. Restricted or local-only context is withheld from cloud inference; this is not offline AI.

### API and webhooks

Tenant-scoped API access and governance-event webhooks. General external-agent governance adapters remain planned.

### Slack and email

Configured governance alerts can reach Slack or email. Conversational agent execution in these channels is not offered here.

### Microsoft Teams

A collaboration surface is proposed. It is not currently an available agent runtime or approval channel.

Complete model-spend accounting remains in progress. The 420 use case explorer entries are proposals, not deployed agents or verified integrations.

Source: https://centerofagentic.com/platform

## Related pages

- [Book a demo](https://centerofagentic.com/contact?intent=demo)
- [Explore use cases](https://centerofagentic.com/use-cases)
- [Set the least authority the work needs](https://centerofagentic.com/platform/governance#authority)
- [Explore the release controls](https://centerofagentic.com/platform/governance#release)
- [](https://centerofagentic.com/images/coa-policy-simulator.webp)
- [Understand the run evidence](https://centerofagentic.com/platform/governance#evidence)