API Drawer: snap your API specs together.
Build something new.
Pair a weather API with a transit API, and you've got a commute bot that tells you to grab an umbrella. API Drawer keeps your APIs in one place, exports the ones you pick as clean API docs, and hands them to your AI through an MCP server.
Let your AI run your API drawer
Add your personal URL to any MCP-compatible AI tool once. After that, just ask: it finds APIs, adds new ones, and saves sample responses. No more copying specs field by field.
Connect in 4 steps
- 1Sign in with Google.
- 2Copy your MCP URL from the Connect AI page.
- 3Add it once to an MCP client like Claude Code.
- 4Ask in plain English to find APIs, add new ones, and save samples.
claude mcp add --transport http apidrawer "<your-personal-url>"
Browse and look up APIs
Your AI sees what's in your drawer, plus each endpoint and its response fields.
Read sample responses
It works from real response examples, starting with your primary sample.
Add APIs
Describe an API in chat and it lands in your drawer, marked as verified.
Save sample responses
It tidies up response examples, masks personal data, and saves them for you.
Sample chat
You: Do I have a weather API in my drawer? If so, what fields does it return? AI: (calls list_apis → get_api) Yes, Hourly forecast. It returns ... You: Add the MTA bus arrivals API. It uses an API key. AI: (calls create_api) Added BUS_ARRIVALS_GET and marked it as verified.
What your AI handles · What you handle
Your AI
- Looks up your saved APIs and samples
- Adds new APIs (saved as verified)
- Cleans up and saves sample responses
You
- Review new APIs in the app
- Switch off APIs you no longer need from the details page
- Build docs yourself on the Export page
From drawer to product in 3 steps
No setup. Collect your APIs, combine them, and turn the combo into something real.
1. Stock your drawer
Save public APIs, internal endpoints, and open-source services as blocks you can pull out anytime.
2. Mix and match
Check off two or three APIs at a time and see what product ideas fall out.
3. Hand it to your AI
Copy the combined spec and prompt in one click. Your AI takes it from there, from product plan to working code.
A combo doc your AI can build from
The APIs you pick merge into one clean spec and a build prompt, ready to paste into ChatGPT, Claude, or Cursor.
# API Reference
Generated: Sep 29, 2026 · 1 API · Source: API Drawer
## 1. Hourly forecast (WEATHER_HOURLY_GET)
- Status: Verified
- Request: GET https://api.example.com/v1/forecast/hourly
- Auth: API_KEY (key saved, value not included)
- Category: Lookup · Tags: weather, public data
- Description: Returns the hourly temperature and chance of rain for a US ZIP code.
### Request parameters
| Name | Type | Required | Description | Example |
|---|---|---|---|---|
| zip | string | Yes | 5-digit US ZIP code | 10001 |
| date | string | Yes | Forecast date (YYYY-MM-DD) | 2026-09-29 |
### Response fields
| Name | Type | Nullable | Description | Example |
|---|---|---|---|---|
| hour | string | No | Forecast hour, 24-hour clock (00–23) | 00 |
| temp_f | number | No | Temperature (°F) | 65.3 |
### Sample response (primary sample: New York City, 1 day)
```json
{
"items": [
{
"hour": "00",
"temp_f": 65.3
},
{
"hour": "01",
"temp_f": 64.4
},
{
"hour": "02",
"temp_f": 63.5
}
]
}
```
Omitted: $.items (21 more)
Not another API tester. A drawer for building.
Most API tools, from API clients and testers to docs portals, help you call and test one endpoint at a time. API Drawer keeps the APIs you use, with field notes and real samples, so you can combine them into whatever you build next.
| Typical API tools | API Drawer | |
|---|---|---|
| What it's for | Call and test one endpoint at a time | Combine several APIs into a product that doesn't exist yet |
| Who writes the specs | Mostly you, by hand or by importing files | Your AI. Connect it once and it fills in every field description |
| Handing off to AI | Paste the entire spec and watch your chat bog down | Only the APIs you pick, with samples trimmed to 3 items |
| What you walk away with | A 200 OK | A product brief your AI can start building from |
What it's for
Who writes the specs
Handing off to AI
What you walk away with
Your keys stay locked down
Your credentials are kept apart from your specs and stored securely.
Encrypted, separate key storage
API keys are encrypted with your own encryption key before they're stored.
Keys never ship with your exports
Markdown and Excel exports leave out every key value, so you can paste them into any AI chat.
AI access through your personal URL (MCP)
Add your personal URL to an MCP-compatible AI tool, and it pulls only the API details and samples it needs, live.
Start your first API combo
Your APIs are scattered across docs, bookmarks, and old projects. Put them in one drawer and see what they make together.