External AI provider · 23 pre-built actions
Anthropic integration with Nexus
Call Anthropic from a workflow when policy allows. Messages, message batches, models, workspaces, users and invites actions let workflows use Anthropic as an optional external system.
This is an external AI provider. It is separate from Nexus Sovereign AI, which runs models inside your deployment. Workflows call this provider only when your policy allows data to leave your environment.
Pre-built actions
What you can automate
23 pre-built Anthropic actions, grouped by what they work on.
23
PRE-BUILT ACTIONS
- 01Messages and Message Batches actions send requests to Anthropic models.
- 02Models actions list available models.
- 03Workspaces, Users and Invites actions administer the Anthropic organisation.
Workflow
How it fits into a Nexus workflow
Anthropic steps run as pre-built actions. A run that uses them starts from a webhook, a schedule, a request in Nexus or another system's trigger source.
A trigger source starts work. A connector action performs work. Both run inside the same governed execution.
Beyond pre-built actions
When the pre-built action is not enough
Move down one step: an API, protocol or database call, then custom logic in Python, Bash, Go, Java, .NET, PHP, PowerShell or JavaScript/TypeScript with any library. Every step uses the same secrets, approvals, history and audit.
01
Call the Anthropic API with the HTTP Request node, or use GraphQL, SOAP or a database node where the system exposes one.
02
Write custom integration logic in Python, Bash, Go, Java, .NET, PHP, PowerShell or JavaScript/TypeScript, with the libraries your team already uses.
03
Direct calls and code use the same vault-backed secrets, approvals, run history and audit as pre-built actions.
Examples
Workflow examples
Policy-gated summarisation
Rules decide whether data may leave the environment before a request is sent.
Batch processing
Large sets of non-sensitive items are processed through message batches.
Account administration
Workspace and user changes follow approval and are recorded.
Deployment, trust and audit
Governance, audit and deployment
Deployment
Nexus runs on-premises, in private or sovereign cloud, or air-gapped. The step needs network reach to the system it calls, and nothing else changes.
Secrets and identity
Credentials are referenced through vault-backed secret handling and role-based access, never embedded in workflow logic.
Audit
Inputs, outputs, approvals and outcomes are written to the execution history for every run.
FAQ
Anthropic connector questions
- Which Anthropic actions are pre-built?
- Nexus includes 23 pre-built Anthropic actions across Workspaces (5), Invites (4), Message Batches (6), Messages (2), Users (4) and Models (2).
- Can Anthropic start a workflow?
- Runs that use Anthropic start from a webhook, a schedule or another trigger source. Anthropic steps run as actions inside the workflow.
- What if the operation I need is not pre-built?
- Use the Anthropic API through the HTTP Request node, or write the step in Python, Bash, Go, Java, .NET, PHP, PowerShell or JavaScript/TypeScript. It runs in the same workflow with the same approvals and audit.
- Is this the same as Nexus Sovereign AI?
- No. Sovereign AI runs models inside your deployment. Anthropic is an external provider that workflows call only when your policy allows.
Related reading
Continue from here
- Nexus Sovereign AI
Models that run inside your deployment.
- OpenAI connector
Another optional external provider.
- Human-in-the-loop
Where approval belongs in a running workflow.
- Integration patterns
How systems work together in a complete process.
Integration review
Discuss how Anthropic fits your process.
Bring the operation you need. We will map it to a pre-built action, or to the API, protocol or code step that covers it.