Platform · Integrations

Connect to the systems
your business already runs on

35+ native connectors, a typed connector SDK, and a credential vault that means agents never touch a secret. Every connector call is audited.

35+Native connectors
4Transport protocols
6AI model providers
SDKFor custom connectors

Connector library

Native connectors are maintained in the platform. “Via Connector SDK” means you can configure them using the same SDK — without custom code.

ServiceNow

ServiceNow

Native

Jira Service Mgmt

Jira Service Mgmt

Native

Zendesk

Zendesk

Native

Freshdesk

Freshdesk

Native

ME

ManageEngine

Native

PagerDuty

PagerDuty

Native

Salesforce

Salesforce

Native

HubSpot

HubSpot

Native

Zoho CRM

Zoho CRM

Native

Pipedrive

Pipedrive

Via Connector SDK

Slack

Slack

Native

Microsoft Teams

Microsoft Teams

Native

WhatsApp Business

WhatsApp Business

Native

Email (SMTP/IMAP)

Email (SMTP/IMAP)

Native

Telegram

Telegram

Via Connector SDK

PostgreSQL

PostgreSQL

Native

MySQL

MySQL

Native

Elasticsearch

Elasticsearch

Native

Confluence

Confluence

Native

SharePoint

SharePoint

Native

Notion

Notion

Via Connector SDK

Google Drive

Google Drive

Via Connector SDK

Okta

Okta

Native

Azure Active Dir.

Azure Active Dir.

Native

LDAP / OpenLDAP

LDAP / OpenLDAP

Native

SAML

SAML 2.0

Native

Google Workspace

Google Workspace

Via Connector SDK

Shopify

Shopify

Native

AWS

AWS

Native

AZ

Azure

Native

Google Cloud

Google Cloud

Native

OPN

On-Premise Servers

Native

Kubernetes

Kubernetes

Via Connector SDK

OpenAI GPT-4o

OpenAI GPT-4o

Native

Azure OpenAI

Azure OpenAI

Native

Anthropic Claude

Anthropic Claude

Native

Google Gemini

Google Gemini

Native

BYOM

Private / Self-hosted

Native

Ollama (local)

Ollama (local)

Via Connector SDK

Need a connector that's not in the library?

The Connector SDK lets you define typed connectors that participate in the same credential vault, audit logging, and permission machinery as native connectors. Build once, govern the same way.

Talk to an engineer
Integration architecture

How integrations actually work

The design choices behind the connector layer — why agents never hold secrets, how permissions are enforced, and what extensibility looks like without compromising governance.

Connection model

Connector, not integration spaghetti

Every integration runs through a typed connector interface. Credentials are stored once, in a vault, and referenced by name. Agents never hold connection strings — they call a named connector. This makes auditing, rotation, and access revocation a configuration change, not a code change.

  • Typed connector spec (source, action, auth method, schema)
  • Credential vault with per-connector access policies
  • Connector health status surfaced in the control plane
  • Zero connector credentials in agent prompt or memory
Permissions

Least-privilege by construction

Each connector declares a capability set — read, write, delete, admin. Agents are assigned connectors with explicit capability caps at deployment time. The platform enforces capability boundaries at the connector layer; agents cannot escalate their own permissions.

  • Capability caps: read / write / delete / admin per connector
  • Agents bound to declared connector list at deployment
  • Runtime permission checks before every connector invocation
  • Audit log entry on every connector call (agent, connector, action, outcome)
Extensibility

Custom connectors without forking the platform

If a system isn't in the native library, the connector SDK lets you define one — in your language, deployed in your environment. Custom connectors participate in the same credential vault, audit, and permission machinery as built-in connectors. You don't compromise on governance to get extensibility.

  • Connector SDK with typed request/response schemas
  • Webhook receiver primitive for inbound event-driven triggers
  • REST, GraphQL, gRPC, and database transport support
  • Custom connectors version-pinned per deployment
Model routing

The model is a connector, not the foundation

LLMs are treated as connectors — not as the platform itself. This means you can swap, pin, or route across models without changing agent logic. A customer support agent can use a fast model for classification and a reasoning model for resolution. The routing policy lives in the control plane.

  • Per-agent model assignment with version pinning
  • Multi-model routing within a single agent workflow
  • Latency, cost, and capability-based routing policies
  • Private / self-hosted model support (Ollama, vLLM, custom endpoints)
Deployment compatibility

Connectors work the same way
regardless of where you deploy

Whether you run Orchestrik on-premise behind your firewall, in a private cloud VPC, or as a managed SaaS tenant — the connector interface, credential model, and audit format are identical. Migrations between deployment modes don't require connector rewrites.

On-PremisePrivate CloudManaged SaaS

What's consistent across deployment modes

  • Credential vault API and secret format
  • Connector capability model (read/write/delete/admin)
  • Audit log schema and retention controls
  • Custom connector SDK version and interface
  • Model connector configuration and routing rules
  • Health monitoring and connector status surface
For integration teams

What's available for technical evaluation

Enterprise evaluations include full access to connector specifications, credential architecture documentation, and the SDK reference. We don't make you reverse-engineer the integration model from a demo.

  • Connector specification reference

    Full typed schema for every native connector

  • Credential vault architecture

    How secrets are stored, rotated, and scoped

  • Connector SDK documentation

    Build and deploy custom connectors

  • Audit log schema

    Field-level specification for all connector audit events

  • Model routing configuration guide

    Multi-model setup, version pinning, fallback policies

See the connector library in a live environment

We'll walk you through the connector model, credential architecture, and custom SDK with a live instance connected to the systems you actually use.