Platform Links
Links allow AI tools to execute operations on remote sources (databases, APIs, services) through a unified interface.
A connector registers itself via the ArangoPlatformLink CRD, declares its capabilities (description, tags, JSON Schema), and processes jobs submitted by AI tools.
Quick Start
- Deploy a link (Helm chart or manual)
- The Link CRD becomes
Readyand appears in/_inventory - AI tools discover the link, read its schema, and submit jobs
- The connector picks up jobs, executes them, and uploads results
Resource Model
Links are not serverless — they run as regular Kubernetes Deployments that you manage. The link pod contains:
- Your link container — your binary that processes jobs
- Integration sidecar — injected automatically by the platform, provides the job queue (MetaStore) and file storage (StorageV2)
The link runs continuously and polls for jobs. It does not run as a sidecar of the caller — it is an independent workload with its own resource limits that you configure in the Deployment spec.
You are responsible for:
- Setting CPU/memory limits on the link container
- Choosing the number of replicas (multiple instances share the job queue safely)
- Monitoring the link’s health
Sections
- CRD Reference — ArangoPlatformLink resource
- API Reference — REST and gRPC endpoints
- Jobs — Job lifecycle and states
- Building a link — How to build your own connector