Step 1
Write the cases
Put one case per row in a CSV or JSON file, or a dataset view. Each column becomes a variable your requests and scripts read.
Your API works. Now test it against every edge case.
Data-driven testing in Postman runs your collection once per row of test data, from CSV and JSON files or databases, so developers, QA teams, and coding agents find the exact input that fails.
Run it in the Collection Runner in the Postman app, or let your favorite coding agents like Claude Code, Codex, and Cursor run it for you with the Postman CLI.
Trusted by teams at
Install the Postman plugin and your agent gets data-driven testing workflows it can run on its own: writing a CSV or JSON file of edge cases, running a collection once per row, building a dataset with SQL views over your files and databases, and gating CI on the rows that fail.
Every command runs through the Postman CLI, the same tool your CI pipeline uses.
Marketplace
get the Postman plugin
claude plugin marketplace add anthropics/claude-plugins-officialclaude plugin install postman@claude-plugins-officialQuery files and databases with SQL, then run a collection once per row
Run a collection over every row of a CSV or JSON file
Gate every PR on your Postman tests
Keep secrets in the Shared Vault, out of committed files
Try these prompts
WHY POSTMAN
No new test code for each case. Keep your cases in a CSV file, a JSON file, or a dataset, and the collection you already have runs once per row.
Keep the cases in a CSV or JSON file and pass it to the run with -d. Each column becomes a variable your requests and scripts read, such as {{tier}}, so adding a case means adding a row.
The collection runs once for every row. When a test fails, the run names the iteration, so your agent knows which row to check. Signed-in runs of a cloud or linked collection upload to Postman with each test’s result and error details.
A dataset turns CSV, JSON, and spreadsheet files, and MySQL, Postgres, SQL Server, or JDBC databases, into tables you query with SQL. A view picks the rows for one scenario and can join a file to a live database.
Each dataset is a .dataset.yaml file that records where the data lives and the SQL for each view, without copying your rows into it. Changes go through review like the rest of your code.
Use a view as iteration data, query the dataset with pm.datasets in pre-request and post-response scripts, including Postman CLI runs, or return its rows from a local mock.
Database credentials can be Shared Vault references, or environment variables in CI, in place of plain text in the dataset file. Connect with a database account limited to read-only access.
Step 1
Put one case per row in a CSV or JSON file, or a dataset view. Each column becomes a variable your requests and scripts read.
Step 2
Run the collection in the Collection Runner or with the Postman CLI. Each row runs as one iteration.
Step 3
A failed test names its iteration, so you know which row to check.
Step 4
Run the same command in your pipeline and fail the build when any row fails.
Data-driven API testing runs the same requests and tests once for each row of test data, so one collection covers many inputs. In Postman, each row is an iteration, and its columns become variables your requests and scripts read. Adding a case means adding a row.
Yes. Install the Postman CLI with npm install -g postman-cli, then run postman init to install Postman's agent skills in your repository. Claude Code, Codex, Cursor, and other agents can then write a CSV or JSON file of cases and run postman collection run with -d, or build a dataset and run one of its views with --iteration-data-dataset and --iteration-data-view.
A data file is a CSV or JSON file with a fixed set of values for a run. A dataset is defined once and reused across collection runs, performance tests, monitors, mocks, and scripts. It can combine files and live databases, and its SQL views pick the rows for each scenario. Data files must be smaller than 5 MB; a dataset has no file size limit. Use a data file for a simple list of cases, and a dataset when you need to filter, join, or reuse the data. See the Datasets docs.
Add the database to a dataset as a source, save a view that selects the rows you want, and run the collection against that view. From the Postman CLI, postman dataset source add connects MySQL, Postgres, SQL Server, or a database with a JDBC driver, postman dataset view create saves the SQL, and postman collection run with --iteration-data-dataset and --iteration-data-view runs one iteration per row. In the Postman web app you can add MySQL and Postgres sources; local files, SQL Server, and JDBC sources need the desktop app or the Postman CLI. Live database sources need a Team or Enterprise plan, and custom JDBC sources need Enterprise.
Check the iteration number. Each row runs as one iteration, and the Postman CLI lists the iteration for every failed test in its failure summary. When the CLI is signed in and the collection is in Postman or linked to it, results upload to the Run results page with each test’s result and error details.
Use Shared Vault references or environment variables in place of literal values. The Postman CLI accepts vault:<KEY> for a database host, user, or password, and env:<NAME> reads the POSTMAN_VAULT_<NAME> environment variable in CI. A literal password is stored as plain text in the dataset file, so connect with a database account limited to read-only access.
Yes. A performance test can read the same CSV or JSON data file, or a dataset view, and spread its rows across virtual users. From the Postman CLI, postman performance run takes --data-file, or --dataset-id with --dataset-view-id, and --dataset-distribution sets how rows map to virtual users: round-robin (the default), fixed, or random. See Performance Testing.
Data files and Datasets are available on Postman Solo, Team, and Enterprise plans. Connecting a live database source needs a Team or Enterprise plan, and custom JDBC sources need Enterprise. See Postman pricing for what each plan includes.