Inspection report
amazon-ads-simulator-mcp
Open source MCP server from automatasleuth. Tool list: code-checked6.
The answer
0
What can it change in your account?
None of the 8 tools on record can change data.
name based Read and write here are inferred from tool names and descriptions, not from a test.
| Area | Access | Write tools | Read tools |
|---|---|---|---|
| Orders | · Orders no tool on record | 0 | 0 |
| Ads | R Ads read only | 0 | 1 |
| Listings | · Listings no tool on record | 0 | 0 |
| Pricing | · Pricing no tool on record | 0 | 0 |
| Inventory | · Inventory no tool on record | 0 | 0 |
| Reports | · Reports no tool on record | 0 | 0 |
| Finance | · Finance no tool on record | 0 | 0 |
Record
The facts on file.
Tools
All 8 tools we found.
Grouped by what they do to your data. The access column shows why we put a tool there.
read Reads data 8
| Tool | Description | Area | Access basis |
|---|---|---|---|
| explain_recommendation | Explain why a recommendation row was increased, decreased, or held. | Other | name based |
| explain_variable | Explain how a simulator variable should be controlled. | Other | name based |
| get_backtest_summary | Return forward-validation backtest counts and a plain-English interpretation. | Other | name based |
| get_current_config | Return the base rule configuration used by the simulator. | Other | name based |
| get_dataset_manifest | Return the loaded simulator dataset source, row counts, expected files, and backtest metadata. | Other | name based |
| get_simulation_summary | Return current action counts, ACoS/TACoS totals, and reason-code counts. | Other | name based |
| list_recommendations | List top simulator recommendations, optionally filtered by action or confidence. | Other | name based |
| run_scenario | Run the simulator with optional rule and forecast overrides. This is read-only and does not change bids. | Ads | name based |
Method
How we checked this server.
We read the tool definitions in the source code. We did not run the server against a seller account.6
Not tested with credentials. We did not connect this server to a seller or ads account. A tool marked read can still have side effects we could not see, and a vendor can add write tools at any time. Check the list again before you grant access.
Areas come from keywords in tool names, as explained on the matrix page.
Notes
Sources for this report
- 1Source repository or product pagegithub.com
- 2Documentationgithub.com
- 3License: GitHub API repo.license (search 2026-10-30)
- 4Authentication: static scan of repo (env names / code patterns)
- 5Pricing: https://github.com/automatasleuth/amazon-ads-simulator-mcpgithub.com
- 6Our MCP inventory of Oct 30, 2026: tool list, access marks, stars and last commit. GitHub repository search (22 queries, unauthenticated) -> shallow clone -> static extraction (extract_tools.py) -> optional runtime tools/list (mcp_probe.py); vendor/official from public docs. access=read|write is a name/description heuristic unless access_basis says otherwise.