Callable by agents. Governed by you.
The AI Frontier extension layer turns the platform from something people use into a governed capability surface that agents can call, external AI ecosystems can reuse, and your teams can trust — built on governed MCP and signed Skills, with SoCode, SoTag and SoClaw on top.
Operate it from where you already work.
The same governed agent — same MCP server, same scopes, same audit trail — wherever you work: in your editor, in Slack, or running on its own. Not a chatbot bolted onto a dashboard; the platform itself, callable.
SoCode
AvailableA downloadable skill plus the governed MCP server, so an agent can operate the platform and answer questions about your data in plain language — without leaving the editor.
- ›Ask about your data in plain language, in-editor
- ›Run governed operations without leaving Claude Code or Cursor
- ›Backed by the same MCP server and scopes as everything else
SoTag
In ProgressThe same governed agent in the channel where your team already coordinates: ask for a metric or run an approved operation — the answer, and the audit trail, live right next to the decision.
- ›Pull a metric or run an approved action inside Slack
- ›The answer and its audit trail sit next to the decision
- ›No dashboard-hopping just to get a number
SoClaw
Early AccessThe autonomous one. It doesn't wait to be asked — it works your high-value, at-risk customers one by one, picks a personal way to win each back or holds off, and proves the real lift against a control group it never touches. Off by default, with spend caps and an instant pause.
- ›Works high-value, at-risk customers one at a time
- ›Chooses a personal win-back or a deliberate hold
- ›Proves real lift vs an untouched control — capped, pausable
Or run it inside the agent your enterprise already uses.
Because it's a standard, governed MCP server, the platform doesn't only run in our agents — it plugs into the enterprise assistants your teams already have. Same tenant key, same scopes, same audit trail.
Claude Desktop
AvailableConnect the governed MCP server straight into Claude Desktop and ask about your members, campaigns and metrics — or run an approved operation — in an ordinary chat.
How it works →Microsoft Copilot
AvailableA declarative agent for Microsoft 365 Copilot: your team queries members, campaigns and metrics inside the Copilot they already use — with the same tenant key, scopes and audit trail as every other call.
How it works →SAP Joule
PlannedBring the same governed tools into SAP's Joule assistant, so operations and merchandising teams reach loyalty and campaign data without leaving SAP.
Learn more →What it's built on: MCP and Skills.
Two building blocks make the platform callable without giving up control — the interface an agent talks through, and the capabilities it's allowed to invoke.
A standard, authenticated way for any agent — Claude, Copilot, M365 — to call the platform's tools. Every call carries a tenant key, is authorized per tool, metered against a budget, and written to the audit trail.
Packaged business tasks with their conditions, validation and boundaries built in — the units an agent actually invokes. Each one is cryptographically signed and integrity-checked before it runs, and sandboxed while it does.
Under it all
One certified semantic layer.
MCP and Skills are only as trustworthy as the numbers underneath them. Both read through one governed semantic layer — every business metric defined and certified once, computed one way. So the figure an agent returns is the figure on your dashboard, and the AI never invents a number.
One certified semantic layer
Defined once · computed one way
MCP
same number
Skills
same number
Dashboards
same number
Autonomy you can govern.
AI-generated commands are validated against a static command tree before they run; third-party Skills are signed, integrity-checked and sandboxed; and every MCP call is authenticated by a tenant key, authorized per tool, metered against a budget, and audited. Metric numbers come from the semantic layer — the AI never invents figures.
Tenant isolation · least privilege · full auditability · SOC 2 Type II · CCPA/CPRA-ready · Azure-native