Platform capabilities
ATFGraph supports more than the hosted first-node workflow. Reach out to AI-LAB to enable any of the capabilities below for your team, or to discuss a deployment that fits your environment. Availability depends on configuration, access and the support needed for your use case.
| Capability | What it enables |
|---|---|
| Kubernetes & Helm | Run the graph service and workers in a Kubernetes environment with a deployment and operations plan. |
| Docker & Docker Compose | Container-based development, evaluation and self-hosting. |
| Databricks | Managed graph and worker hosting, node deployment and Databricks integrations. |
| HTTP API | Integrate runs, sessions, memory, knowledge and governance into another application. See the API contracts. |
| Client SDKs | Connect applications to the graph using supported clients. AI-LAB can confirm language support and provide the right package and access. The Python SDK is documented here. |
| MCP, A2A & discovery | Expose or discover capabilities through supported protocols and controlled publication. |
| Schedules, webhooks & event sources | Trigger nodes from time-based jobs, authenticated webhooks or connected systems. |
| Coding-agent harnesses | Delegate bounded work to an installed coding-agent CLI within a recorded traversal. |
| Model providers & multimodal capabilities | Configure approved models, tool calls, image/audio/file workflows and realtime features. |
| Knowledge & retrieval | Enable document ingestion, search, retrieval backends and the access required for your data. |
| Identity, access & approvals | Configure identities, credentials, connector grants and durable human review. |
| Observability & MLflow | Connect recorded runs, node logs, usage, tracing and external observability tools. |
| Serverless & on-demand workers | Explore bounded invocations and workers that start when needed. |
| Operations & lifecycle | Plan retention, backups, migration, runtime settings and service monitoring. |
Ask AI-LAB to enable a capability
Share your team and environment, the capability you need, the application or data it will connect to, expected usage and any access requirements. For self-hosting, include the intended Docker, Kubernetes or other hosting environment. AI-LAB can confirm prerequisites and the supported setup before you build against it.
Use your existing AI-LAB contact or internal support channel. See Get support for the details to include. You do not need to install or administer a graph server to follow the hosted quickstart.