> ## Documentation Index
> Fetch the complete documentation index at: https://docs.reeva.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Link engineering docs across Teamcenter and SAP

> Build an agent that links Teamcenter drawings and specs to the right SAP material masters, automatically.

This is the canonical Reeva workflow: keep engineering documents and the data on them **linked and consistent** across Teamcenter and SAP, so engineers stop rekeying and operators stop hunting for the "real" version.

## What this agent does

When a new revision of a drawing or spec is released in Teamcenter:

1. Fetch the document and any related metadata
2. Use AI to match it to the corresponding SAP material master
3. Extract the attributes that belong on that material (grade, dimensions, tolerances, classification)
4. If the match and the extraction are high-confidence, write the link and updated attributes back to SAP
5. Otherwise, flag it in **Monitor** for an engineer to review

## Before you start

* [Teamcenter](/integrations/teamcenter) is **Connected**
* [SAP](/integrations/sap) is **Connected**
* You know which Teamcenter item types correspond to which SAP material types in your business
* A test item in Teamcenter you can use to verify the agent works end to end

## Build the agent

<Steps>
  <Step title="Create the agent">
    On **Agents**, click **New Agent**. Name it something like *"Link drawings to SAP materials"* and add a one-line description.
  </Step>

  <Step title="Add the trigger">
    Add a **trigger** step:

    * Source: **Teamcenter**
    * Event: **New item revision released**
    * Scope: the item types you want to handle (e.g., `Drawing`, `Specification`)
  </Step>

  <Step title="Fetch the document">
    Add an **action** step:

    * Integration: **Teamcenter**
    * Operation: **Fetch dataset files**
    * Output: the attached PDF, native CAD, or spec sheet
  </Step>

  <Step title="Match it to a SAP material">
    Add an **AI** step:

    * Input: the Teamcenter item ID, name, and the document fetched in the previous step
    * Task: *"Find the SAP material master that corresponds to this Teamcenter item. Use the part number, description, and any classification metadata."*
    * Output: a SAP material number and a confidence score
  </Step>

  <Step title="Extract attributes from the document">
    Add another **AI** step:

    * Input: the document content
    * Task: *"Extract the following attributes: material grade, primary dimensions, tolerance class, surface finish."*
    * Output: a structured record and a confidence score per field
  </Step>

  <Step title="Add the confidence gate">
    Add a **decision** step:

    * Condition: match confidence ≥ 95% **and** all extracted attributes ≥ 90%
    * **High-confidence branch:** continue to write step
    * **Low-confidence branch:** mark as **Flagged** in Monitor
  </Step>

  <Step title="Write back to SAP">
    Add an **action** step on the high-confidence branch:

    * Integration: **SAP**
    * Operation: **Update material classification + link document**
    * Inputs: the material number from step 4, the attributes from step 5, the Teamcenter document URL
  </Step>

  <Step title="Turn it on">
    Toggle the agent **on**. Release a test revision in Teamcenter to confirm the agent picks it up.
  </Step>
</Steps>

## Verify it's working

* Open **Monitor** and filter by this agent
* Newly processed items should appear within a minute or two of the Teamcenter release
* Flagged items should be the ones where the match or extraction was uncertain — review them and either approve, edit, or update the agent's instructions

## Tips

* **Run in shadow mode first.** For the first week, replace the SAP write step with an export to **Monitor** only. Review the proposed links and attributes before letting the agent write to SAP.
* **Be strict with your confidence threshold.** SAP material data is hard to fix once it's wrong. 95%+ for the match and 90%+ per field is a good starting point.
* **Use the audit trail.** Every write Reeva makes to SAP is logged. If something looks off in SAP, the [audit trail](/administration/users-and-audit) tells you which agent run produced it.

## Related

* [Route engineering changes](/use-cases/change-orders) — handle engineering change notices once the link exists.
* [Monitor](/agents/monitor) — review what your agent produced.