# @kurlo/cli [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/stormhop/kurlo  
**GitHub Stars:** 58  
**npm Downloads (last month):** 2098  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/kurlo-cli

## Description
Kurlo's command line: run API collections in CI and serve them to AI agents over MCP.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "cli": {
    "command": "npx",
    "args": ["-y","@kurlo/cli"]
  }
}
```

## Documentation

## What @kurlo/cli does

@kurlo/cli provides two command-line workflows for Kurlo workspaces. The `run` command executes API requests and their scripts and tests for CI. The `mcp` command exposes a workspace to AI agents through the Model Context Protocol.

Kurlo workspaces are stored as YAML in a Git repository. The CLI can run saved requests and collections with configured authentication, environments, variables, pre-request scripts, and test scripts. It also supports one-off requests using the workspace's `{{variable}}` values.

The MCP interface lets an agent inspect collections and environments, execute requests or complete collections, create and modify requests and `pm.*` tests, save responses as named examples, and set environment variables. It can read the most recent request sent from the Kurlo app, allowing an agent to investigate a failure without requiring the response to be copied into chat.

## How it works

Run the package with `npx -y @kurlo/cli`, or use a prebuilt CLI binary for macOS, Linux, or Windows on x64 and arm64. A CI run points at a workspace directory and can select an environment, for example:

```bash
npx -y @kurlo/cli run ./workspace --env CI
```

The runner supports data-driven iterations, pretty, JSON, and JUnit output, variable export, and non-zero exit status when execution fails. Collection load-test commands can list load tests, export k6 scripts, or run local k6 tests with threshold results and reports.

Start the MCP service with the `mcp` command. Kurlo handles the request's configured authentication, scripts, tests, and variables while keeping secret values out of model-visible results. Environments marked protected cannot be accessed by the agent.

## Setup and configuration

Install or invoke the npm package directly:

```bash
npx -y @kurlo/cli --version
```

To configure MCP clients automatically on the machine, run:

```bash
npx -y @kurlo/cli mcp install --env Staging
```

Use `--project` when the configuration should be written to repository files, including `.mcp.json` and `.cursor/mcp.json`, so collaborators receive the setup when they clone the project. The desktop application is not required for the CLI or MCP service.

Workspaces can define multiple environments and use `{{variable}}` templates in URLs, headers, parameters, bodies, and authentication fields. Secrets remain local to the workspace configuration and are sent to the target API when needed, but appear as `[secret]` in information returned to the model.

## Tools and capabilities

- Run saved requests and complete collections with authentication, scripts, tests, and environment variables.
- Send one-off API calls using workspace variable substitution.
- List collections and environments for agent-driven API work.
- Add or edit requests and their `pm.*` tests in the workspace YAML.
- Save responses as named examples and update environment variables.
- Execute CI runs with iteration data, selectable reporters, variable export, and failure exit codes.
- List, export, and run collection load tests through k6 workflows.

## Limitations and notes

The MCP service respects protected environments and does not expose their contents to agents. Secret values may be used for API authentication, but are masked in model-visible output. Repository edits are made in the same YAML workspace format used by Kurlo, so the workspace remains suitable for version control and pull-request review. The CLI is distributed separately from the desktop application, and the desktop app is not needed to run it.

_Full upstream README: https://allmcps.com/mcp/kurlo-cli/readme_