Tiger

Use case · MCP API client

Let AI run your
API requests

Tiger ships a built-in Model Context Protocol server. Point any MCP-compatible AI assistant at your collection and it can list, inspect and execute your requests without leaving its own interface.

Download Tiger, free Compare with Postman

Free and open source (MIT)

Background

What is MCP?

Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools in a structured, auditable way. Instead of asking an AI to guess how to invoke something, you expose a typed set of actions it can call directly.

Workflow

How it works in practice

Three steps from zero to an AI assistant that can explore and run your API collection autonomously.

Open your collection in Tiger

Clone the repo that holds your .tiger files, or import an existing Postman export. The MCP server starts automatically when Tiger launches; no separate process or configuration needed inside the app.

Add Tiger to your AI client's config

Drop a single JSON block into your Claude Desktop, Cursor or other MCP-compatible client configuration. Point it at Tiger's compiled server file and the collection path you want to expose.

Ask your assistant to work with the API

The AI can now list every request in the collection, read the full definition of any request including headers, body schema and variable references, and fire a request against your target environment with a single tool call.

Results flow back into context

Response status, headers and body are returned directly to the AI's context window. It can interpret the response, chain follow-up requests, or surface errors without you switching windows.

Configuration

Claude Desktop config snippet

Add the following block to your claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json). Adjust the two paths to match your machine.

The command field points to the Node.js binary on your machine. The first argument is the path to Tiger's compiled MCP server module (inside the app bundle or the project's out/mcp/ folder). The second is the path to the Tiger collection directory you want the AI to access.

Secrets held in environment files are redacted before values are returned to the AI. The AI sees variable names, not secret values. If you want to allow the AI to actually send requests, set allowRun to true in the args.

Restart Claude Desktop after saving. The Tiger tools will appear in the model's tool list on the next conversation.

claude_desktop_config.json

{
  "mcpServers": {
    "tiger": {
      "command": "node",
      "args": [
        "/Applications/Tiger.app/Contents
  /Resources/out/mcp/server.mjs",
        "/path/to/your/collection"
      ]
    }
  }
}

MCP tools

The four tools Tiger exposes

The MCP server registers exactly four tools. Each has a typed input schema so the AI model can call them without guessing argument names or shapes.

Tool name	What it does	Returns
list_requests	Returns every request in the collection grouped by folder, including name, method and path.	JSON array of request summaries
get_request	Returns the full definition of a single request: method, URL, headers, body, auth config and variable references. Secret values are redacted.	Full request object (secrets stripped)
run_request	Executes a request against the active environment and returns the response status, headers and body. Honors variable substitution and auth flows.	HTTP response (status, headers, body)
list_environments	Lists the environments available in the collection with their non-secret variable names, so the AI can confirm which environment is active before running requests.	Array of environment names and variables

Security

What the AI can and cannot see

The MCP server is designed around least privilege. Here is exactly what is and is not exposed.

Secrets are redacted

Variables prefixed with a tilde (~) in Tiger environment files are treated as secrets. Their values are never sent to the MCP client. The AI sees the variable name, confirming the slot exists, but cannot read the value.

Read and run only

The four tools cover reading collection structure and executing requests. There are no tools to write, delete or modify requests. The AI can explore and invoke, but cannot change your collection files.

Entirely local

The MCP server process runs on your machine. No data is sent to any Tiger cloud service because there is no Tiger cloud service. Communication is only between your AI client and the local Node.js process.

Requests go to your APIs

When run_request executes a call, the HTTP request travels directly from your machine to your target API, exactly as if you had pressed Send in the Tiger UI. No proxy, no intermediary. TLS, client certificates and proxy settings from Tiger's configuration apply as normal.

Scope-limited by path

The collection path you pass in the config determines the scope. The MCP server can only access requests inside that directory. Other collections or files on your machine are not reachable.

Honest caveat

Allowing an AI to execute real API requests carries the same risk as allowing any tool to do so. Review which environment is active before pointing an AI assistant at a production collection. Use a staging or sandbox environment for exploratory AI workflows.

Tiger's MCP server does not add authentication between the AI client and the server process. Anyone who can reach the local port can call the tools. This is standard practice for local MCP servers today, and consistent with how all other MCP tools in the ecosystem operate.

Competitive context

Why no other client ships this

Cloud-based API clients cannot safely expose an MCP server because their collections live in a vendor cloud behind authentication. A local process with read access to your collections only makes sense when the collections are local. Tiger's git-native, file-based approach is what makes the MCP server practical.

Client	Collections stored locally	MCP server included	AI can run requests locally
Tiger	YesNEW	Yes	Yes
Postman	Cloud-primary (local collections on paid tier)	No	No
Bruno	Yes	No	No
Insomnia	Cloud-primary, local mode available	No	No

Table reflects publicly documented features as of June 2026. If any entry is wrong, let us know.

Get started

Give your AI assistant
real API access

Download Tiger, open your collection, add the config snippet, and your AI client can start listing and running requests in under five minutes.

Download Tiger, free Migration guide