We're building the layer between AI and your ERP that should have existed all along.
ERP systems run the world's businesses, but they were never built for AI to safely act inside them. So most teams either bolt AI on as a chatbot, or spend six months engineering a fragile custom integration with no approval gates, no audit trail, and no plan for the next model that ships. We're building DataTether so neither has to be the answer.
Talk to the teamDataTether — live governance
Every AI action, under control
Your ERP — unchanged, unmoved
Scoped tools · Live dashboards · Audit log
Named human reviews the payload before any write runs
ChatGPT · Claude · Claude Code · Manus · Any MCP agent
Origin
The integration project that wouldn't end.
We kept seeing the same project in every ERP team we talked to. A request comes in: "let our AP team ask Claude about overdue invoices." Six weeks later they have a half-built MCP server, no field-level scopes, no approval flow, no audit log — and a new request to also support ChatGPT.
The MCP protocol solved the wiring problem. It didn't solve the harder one: what happens when an AI assistant can actually do something inside a system of record? Who decides what it sees? Who confirms a write? Who's accountable when something changes at 2am?
We built DataTether around a different starting point. Connect once. Govern everything. Let the AI act — but only inside the lines your team draws. And do it without writing a server, hosting infrastructure, or rebuilding every time a new assistant comes out.
Who this is for
Your role. Your exact problem. How DataTether removes it.
ERP / SAP Lead
"One connection, not another six-month project."
ERP / SAP Lead
The person responsible for keeping ERP systems stable and productive
How DataTether helps
Connect once and every AI tool is generated automatically. No custom server to build, host, or maintain.
Head of IT / CIO
"AI productivity without production risk."
Head of IT / CIO
The executive accountable for every live system in the business
How DataTether helps
Scoped access and named-approver gates mean AI value reaches the business without uncontrolled access to live data. No incident. No remediation budget.
Compliance / SOX Officer
"Every change has a name, a reason, and a timestamp."
Compliance / SOX Officer
The person auditors contact first after any system change
How DataTether helps
Every AI-initiated write records the approver, time, and business reason automatically — audit-ready on day one, no manual reconstruction needed.
AI / Automation Lead
"One integration that outlives every model we adopt."
AI / Automation Lead
The team connecting AI assistants to the rest of the organisation
How DataTether helps
One ERP configuration works with ChatGPT, Claude, Manus, and any new assistant that ships. Adopt new AI tools without touching what you already built.
What we believe
Four convictions we won't compromise on.
Every decision we make about the platform (what to ship, what to gate, what to leave to your team) traces back to one of these.
AI earns its place when it can act, safely.
Answering questions is table stakes. Real value shows up when AI can move money, post a journal, or approve a PO inside the system of record — without the team losing control of what it's doing.
The ERP is the source of truth. Always.
No migration. No second database. No "AI-friendly copy" of your data. Your ERP stays exactly where it is, and DataTether works on top of it.
Governance is the product, not a feature.
Scope, approval, and audit aren't a checkbox tier. They're the default behaviour, on day one, for every connection — because the cost of getting this wrong at an enterprise is unrecoverable.
Model-agnostic, by design.
The model your team uses today won't be the model your team uses in 18 months. Build once. Reuse across ChatGPT, Claude, Manus, and whatever ships next.
Our technical bet
MCP-native from the protocol up.
Most platforms treat MCP as a connector plugin, a late addition bolted onto an existing ERP product. They expose the entities their vendor decided to expose, on their vendor's release cycle, with limited configurability. DataTether is built on the protocol itself. It works with any OData service (SAP, Odoo, Oracle, Dynamics, or any custom ERP API) without locking you to one vendor's data model.
When you configure an entity, DataTether generates the complete MCP tool definition on the fly (typed parameter schemas, field-level permissions, operation scope) in milliseconds. Change your configuration and the definition rebuilds immediately. No developer needed, no deployment step, no rebuild cycle.
At runtime, every AI request flows through the DataTether backend: identity is validated, service ownership confirmed, credentials resolved from the vault, and the call translated into the precise OData operation — before a single byte reaches your ERP. Approval gates are enforced at this layer. So is the audit log.
Protocol stack
Your AI clients
ChatGPT · Claude · Any MCP agent
MCP protocol — DataTether native layer
Generates typed MCP definitions from your config — instantly on every change
Routes every AI call: validates auth, resolves credentials, assembles the OData request
Enforces approval gates and writes the audit record before any ERP write executes
Any OData service
SAP · Odoo · Oracle · Dynamics
Why DataTether
What you get — and what you stop carrying.
The work of connecting AI to an ERP doesn't disappear — it moves to us. Here is what that means for your team.
For the ERP / SAP lead
You stop re-building the same plumbing.
Every enterprise builds the same MCP server, auth wiring, and approval flow from scratch — then owns it forever. You skip all of it. One connection, tools auto-generated, nothing to host or patch.
For the CIO and AI lead
The protocol stays current, without your team chasing it.
MCP is young and still moving. Left to you, every release means another round of implementation, testing, and redeploy, or staying stranded on an old version. We keep you on the current protocol so that work never lands on your roadmap.
For the budget owner
A world-class MCP, without hiring scarce GenAI talent.
Senior gen-AI engineers are hard to find and a six-figure decision to hire. We close that gap: you get a governed, production-grade MCP for your enterprise without building a team to make and maintain it.
MCP is early — that's our job, not your risk
Adopting an emerging standard inside an enterprise normally means a recurring bill: change management, regression testing, and re-rollout to a lot of people every time the protocol shifts. We absorb that churn across every customer and use case — so you get the upside of being early without the cost of keeping up.
Honest status
We are pre-launch. Here's what that means for you.
A pre-launch product is a different conversation than a mature one. We'd rather you start that conversation with the full picture than the polished one.
Where we are
Private beta with a small number of partners. SAP, Odoo, and Dynamics services connected and running.
What's working today
OData auto-discovery, generated MCP tools, live dashboards, named-approver gates, audit logging, vault-stored credentials.
What we won't pretend
No customers we don't have. No logos we haven't earned. If the problem we describe matches a problem you have — that's exactly who we built this for.
Talk to the team
Bring one ERP service. Help us build what your team actually needs.
If your team is asking "how do we let AI help, without losing control of the system of record?" — that's the conversation we want to have. Partners who join now work directly with us to shape the platform. Your workflows, your edge cases, your priorities decide what ships next.
Your priorities shape what ships next
Partners work directly with our engineering team. The features you need first get built first — your reality drives the roadmap.
Your real service, not a sandbox
We connect to a service that matters to your team and build the governed surface on top of it — together.
SLA and support that grows with you
From priority response times to a dedicated engineer who knows your setup — your support relationship scales as you do.
Preferential terms, locked in early
Partners who join now get preferential pricing when general availability launches. The terms you agree to today are the terms you keep.
One conversation tells you whether the fit is real. If it is, you help design what comes next.