Docs: connect, govern, and ship AI on your ERP
Everything you need to connect an OData service, generate MCP tools, build review surfaces, and let AI act behind human approval. New here? Start with Getting started, then follow the path through Connections, Advanced UI, and Code usage.
New to DataTether?
Bring one OData service and leave with a governed AI answer your team can inspect.
Understand MCP
- What is MCP?Why AI can answer questions but not safely act — and what MCP changes.
- Tools, prompts & resourcesThe three building blocks that turn AI that talks into AI that does the work.
- ArchitectureWhere the safety, speed, and audit trail come from — and where DataTether fits.
- MCP vs. direct API keysHanding an AI raw credentials vs. giving it a supervised, audited door.
Build and expose
Use it
Frequently asked questions
If you have questions before turning OData services into governed MCP tools, start here.
What is an OData to MCP implementation checklist?
An OData to MCP implementation checklist is a guided way to choose the right service, map metadata into tools, set access controls, preview generated surfaces, and test workflows before production rollout.
What should we scope before exposing an OData service to AI?
Start by deciding which entity sets, fields, functions, and write operations the model should never see. A narrow first scope makes review, governance, and user validation much easier.
How should SAP OData operations be named as MCP tools?
Rename technical OData operations into business-readable tool names that describe intent, such as list open orders or update customer status. Clear naming helps AI clients choose the right action and helps operators review what will happen.
What should admins preview before publishing MCP tools?
Admins should inspect generated tool schemas, UI resources, prompts, exposed fields, confirmation rules, and authentication behavior before making the MCP endpoint available to users or AI clients.
How do SAP OData MCP tools work together?
SAP OData MCP tools expose approved SAP OData endpoints as callable actions so agents can read data, trigger workflows, and return structured results using governed business rules and permissions.
Can MCP tools safely write back to SAP through OData APIs?
Yes, when write operations are explicitly enabled and guarded. Safe writeback usually includes role-based permissions, validation rules, human confirmation steps, and traceable logs before records are created or updated.
How do you secure an SAP OData MCP integration?
Secure implementations use least-privilege access, scoped credentials, endpoint allowlists, audit logging, environment separation, and approval checks for sensitive actions.
What is the best way to roll out OData workflows in production?
Start with read-only workflows, validate outputs with real users, then add guarded write actions where the process and approval path are clear. This phased rollout reduces risk before wider adoption.
How much internal IT effort is required for a first workflow?
Internal effort is usually focused on access, security review, and stakeholder signoff. Keeping the first workflow narrow helps reduce coordination overhead and keeps delivery predictable.
What happens after the first workflow is live?
Once the first workflow is validated, teams can reuse the patterns for tool naming, permissions, UI review, and confirmation gates to prioritize additional SAP or OData-backed workflows.
Follow the implementation path
From first connection to governed AI actions in one afternoon, no custom code.
Docs to production in a single session.
Register your OData service, review the auto-generated tools, build a live dashboard, and gate your first write behind human approval. Every step in the docs maps to a real console action.
Months of dev time eliminated
Auto-generated tools replace weeks of custom MCP server development.
No infrastructure to run
Fully managed. Nothing to install, host, or maintain.
Instant dashboards and UI
Live tables, KPIs, and review screens — no frontend developer needed.
One setup, every AI client
ChatGPT, Claude, Manus, custom agents — all from the same connection.
A good first step is small: one service, one user outcome, one visible control boundary.
Request guided access