Integrations
Additive to the stack you already run
Agent Trust Cloud orchestrates and enriches existing security systems rather than duplicating them. Connectors borrow the signals your identity provider, cloud, SIEM and model platforms already produce, and return decisions those systems can act on.
Build order
Secure the point of creation first, then the point of consequence
Prioritize integrations by control value, installed base and ability to produce differentiated agent visibility. Avoid chasing logos before the platform has repeatable connector mechanics.
Live today: discovery connectors for Microsoft Entra ID, Slack and MCP configs, and the OpenShell policy bridge for agents that run in an OpenShell sandbox. The waves below are the wider plan.
Wave 1
Where agents are created and authenticated
AI & model
- OpenAI
- Anthropic
- Azure OpenAI
- Google Gemini / Vertex
- AWS Bedrock
Identity, cloud & dev
- Microsoft Entra
- Okta
- AWS
- Azure
- Google Cloud
- GitHub
Tooling
- MCP gateway & servers
- Generic HTTP / OpenAPI proxy — Planned
Wave 2
Where agents take business action
Enterprise SaaS
- Microsoft 365
- Google Workspace
- Salesforce
- ServiceNow
- Slack
- Teams
Wave 3
Where the consequences are largest
ERP, PAM & SIEM
- SAP
- Oracle
- Workday
- CyberArk
- Splunk
- Microsoft Sentinel
Waves reflect control value and installed base, not logo count. Later waves are sequenced by customer demand. Where they sit against everything else being built is on the product roadmap.
Connector framework
Integration Architecture
Integration breadth is a major moat, but connectors can also become a maintenance burden. Build a connector framework with shared authentication, pagination, rate-limit handling, schema mapping, health monitoring and versioning so each integration is not a one-off project.
What we build
- Connector SDK/runtime for pull scans, webhooks, event streams and action brokering.
- Credential abstraction supporting OAuth, service principals, workload identity, API keys and certificates.
- Canonical mapping layer for agents, tools, identities, resources and actions.
- Connector health dashboard with sync lag, failures, scopes and last success.
- Permission minimization guidance for each connector.
How it works
- Separate read/discovery scopes from write/enforcement scopes.
- Use idempotent sync and checkpointing for large inventories.
- Test against provider sandbox/dev tenants where available.
- Version vendor schemas and map breaking changes before rollout.
SCALABILITY
The product should be able to add the 50th connector faster and more safely than the 5th.
Need one we do not have?
Use the live policy API while generic proxying is planned
The developer platform exposes live pre-action policy checks directly. The Generic HTTP / OpenAPI proxy is Planned and does not protect tools today; use the policy-decision API or a shipped connector for a pilot.