Data Load Impact turns a planned Salesforce bulk data load into a ratified pre-load safety package, an optional small-batch sandbox rehearsal, and an ordered backlog for your team. It analyzes the load and never performs the production load or changes Salesforce metadata.

Data Load Impact is an Elements MCP playbook. See What is an MCP playbook? to learn how playbooks work.

What it does and when to use it

Use the playbook to understand what a planned load will fire, choose load parameters, identify safety controls, or plan a backfill rehearsal.
It predicts the transitive automation and locking effects of the specific load, audits bypass options, records the data-load owner's safety rulings, and prepares an ordered backlog.
The owner sees reasons to rehearse and chooses whether to use the recommended full-copy sandbox rehearsal. If rehearsal is skipped, the package states the risk.
The playbook analyzes only. Your team applies checklist changes and executes the production load and backlog.

Requirements

Install and connect the core elements plugin to the Elements MCP server. Core Metadata Management subscribers also need the Automation & Execution Intelligence add-on. See Connect your AI assistant to the Elements MCP server for core setup.
Access to the run_data_load_impact, list_reference_models, get_space_context, list_metadata_by_criteria, search_metadata, get_metadata_component, get_metadata_source and explain_metadata_components tools.
Python 3 using the standard library only, and a host that runs the shipped local scripts, such as Claude Code or the Claude app.
The script host needs network access to Amazon Simple Storage Service (S3) for source downloads.
A named data-load owner, a planned source path or Salesforce Object Query Language (SOQL) file, and local access to the source files.
A full-copy sandbox when the owner chooses rehearsal. The Salesforce command-line interface (sf CLI) signed in to that sandbox is optional.
In the Claude app, an attached file may be read into the conversation. For a file with real rows, use Claude Code or attach a copy containing only the columns the load writes.

Install

Claude Code

/plugin marketplace add Q9Elements/elements-mcp-server
/plugin install elements-data-load-impact@elements

Codex

codex plugin marketplace add Q9Elements/elements-mcp-server
codex plugin add elements-data-load-impact@elements

How it runs

Kickoff: the owner provides their name and local source path. The playbook creates or resumes the engagement in the session working directory.
Prerequisites: the playbook confirms the Elements connection, resolves the active Space and reference model, and checks required tools and source access.
Intake manifest: the owner confirms the proposed object, operation, counted rows, written fields, loader, Application Programming Interface (API), and applicable upsert or portal-role facts in one reply. For update or upsert, you run the generated SOQL query and save its result locally.
Impact analysis: the playbook analyzes the confirmed load and predicts its transitive automation and locking effects.
Bypass audit: the playbook reviews downloaded metadata source files locally and records evidence for whether each predicted automation fires.
Safety package ruling: the owner accepts or changes proposed actions and replacement checks, answers the incident-class question, and gives a go or no-go decision with any stated risks.
Optional rehearsal: the owner chooses whether to rehearse in a full-copy sandbox. The playbook can slice up to 200 rows from a local comma-separated values (CSV) source. With a signed-in sf CLI that can match the ratified parameters, it runs the small test batch after you confirm the sandbox checklist is applied for that CLI user. Without the CLI path, you run the batch manually and supply same-job debug logs and loader success and error files.
Backlog: after rehearsal is closed or skipped, the playbook delivers the ordered stories and final report.

What you get

report/final-report.md, the final safety and impact report.
Five or six ordered Markdown story files under stories/ for your team to execute.

What stays with you

In Claude Code, raw CSV rows, SOQL results and debug-log lines stay on your machine, and the session receives only local script outputs with metadata names, aggregates and counts. A file attached in the Claude app may be read into the conversation (see Requirements).
You apply checklist metadata changes, run any manual rehearsal batch, and execute the production load and backlog.
The playbook never performs the production load or changes Salesforce metadata.

Tested with

Claude: command-line interface (CLI) and the Claude app.