Name: GitHub Tools API: Curated Open Source Tools and AI Models
Brand: myworkingmemory
SKU: api-myworkingmemory-software-library-search
Availability: InStock
Author: API Market

Myworkingmemory LLC

# Curated Finds API

## Overview

The Curated Finds API gives you read access to a hand-curated collection of open source tools from GitHub and local AI models from Hugging Face, picked for everyday usefulness: tools that mean less to remember, fewer repetitive steps and simpler ways to get things done. It has 3 endpoints. Search reads the collection, either all of it (leave `query` empty) or filtered by a free-text purpose, a category and a platform, with `limit` and `offset` for paging. Each result carries an `id` such as `github:laurent22/joplin`, and the records endpoint returns that one record when you pass the `id` back as a query parameter. The catalogue endpoint returns the collection version, its latest review date, the number of records and the full lists of category and platform values you can filter on. The API describes software; it does not install or run any of the tools or models it lists.

It is built for developers adding a software discovery or "tools for this task" feature to an app, newsletter or internal portal, for people who keep a personal list of open source tools and local AI models, and for AI agents that need a vetted list of tools with source links to suggest to a user. Every record has the same shape: `id`, `name`, a plain-language `purpose`, `category`, `type`, `platforms`, a `sourceURL` to the repository, a `getSoftwareURL` for downloads or setup, the declared license with a `licenseURL`, any `requirements`, an `evidenceLimits` list that says what was and was not checked, and the `reviewedAt` date. Search responses wrap records in an envelope with the collection `version` and the `total` number of matches, so you can page through results and detect when the collection changes.

## Features

*   **3 endpoints**: Search the collection, get one record by id, and read the catalogue summary with all filter values.
*   **Simple inputs**: An optional free-text `query` (up to 200 characters), optional `category` and `platform` filters, `limit` from 1 to 25 and `offset` up to 10,000. An empty query reads the whole collection.
*   **Source and setup links**: Every record links to the source repository and to a download, install or setup page, with the declared license and a link to the license file.
*   **Clear evidence notes**: Each record states how it was reviewed in `evidenceLimits`, for example "Source review only; software was not installed or run". Records come from individual review of public software sources and documented install routes.
*   **Requirements up front**: The `requirements` and `platforms` fields say what hardware, operating system or setup a tool or model needs, such as an Apple Silicon Mac or a browser with WebAssembly.
*   **Versioned collection**: Every search and catalogue response carries a `version` hash, and the catalogue gives the latest `reviewedAt` date and the record `count`.
*   **Rate limit**: 10 requests per second on both the Free and Pro plans.

## Use Cases

*   **Software discovery features**: Show users open source tools for a task, such as notes, focus or file backups, filtered to the platform they use.
*   **Local AI model finder**: List Hugging Face models and local AI tools with their hardware requirements so users can see what will run on their machine.
*   **AI agent tool suggestions**: Give an agent a reviewed list of productivity tools with source links and setup pages to recommend, instead of guessing from the web.
*   **Curated newsletters and digests**: Page through the collection and pick new finds by `reviewedAt` date for a weekly roundup.
*   **Internal tool directories**: Feed an approved-software portal with names, purposes, licenses and install links.
*   **License checks**: Read `licenseDeclaration` and `licenseURL` for a tool before recommending it inside a company.

## API Documentation

### Base URL

All requests go through the API.market gateway at this base URL:

```
https://prod.api.market/api/v1/myworkingmemory/software-library-search
```

### Authentication

Every request needs your API.market key in the `x-api-market-key` header:

```
x-api-market-key: YOUR_API_KEY
```

Replace `YOUR_API_KEY` with your actual API key.

### Endpoints

#### 1. Search the collection

Returns records from the collection that match a free-text purpose, a category and a platform, with the collection `version` and the `total` number of matches. Leave `query` empty to read the whole collection page by page. Use `limit` and `offset` to page: with `limit=25`, the second page is `offset=25`. Some records have an empty `purpose` or `requirements` string, so treat empty strings as "not provided".

**Method:** `GET`

**Endpoint:** `/search`

**Parameters:**

| Name | In | Description | Type | Required | Default | Max Value |
|---|---|---|---|---|---|---|
| `query` | query | Free-text description of what you want the software to do, for example `notes`. Leave empty to read the whole collection. | string | No | - | 200 characters |
| `category` | query | Category to filter on. Use a value from the `categories` list returned by `/catalogue`, for example `Remember & organize`. | string | No | - | 100 characters |
| `platform` | query | Platform to filter on, for example `macOS`, `Windows`, `Linux` or `Android`. The full list is in `/catalogue`. | string | No | - | 100 characters |
| `limit` | query | Number of records to return, from 1 to 25. | integer | No | `10` | 25 |
| `offset` | query | Number of records to skip, for paging. | integer | No | `0` | 10000 |

**Example Request:**

```bash
curl -X GET "https://prod.api.market/api/v1/myworkingmemory/software-library-search/search?limit=2&offset=0" \
  -H "x-api-market-key: YOUR_API_KEY"
```

**Example Response (200 OK):**

```json
{
  "version": "c8f943dfc40e07f096a1fc0ae00040dc14d1df32657e8ae91f82d01aed3fab9b",
  "total": 227,
  "records": [
    {
      "id": "github:activitywatch/activitywatch",
      "name": "ActivityWatch",
      "purpose": "Automatically keeps a timeline of active apps, browser tabs, and idle periods so you can review where computer time went.",
      "category": "Focus & pacing",
      "type": "app",
      "platforms": [
        "macOS",
        "Windows",
        "Linux"
      ],
      "sourceURL": "https://github.com/ActivityWatch/activitywatch",
      "getSoftwareURL": "https://activitywatch.net/downloads/",
      "licenseDeclaration": "MPL-2.0",
      "licenseURL": "https://github.com/ActivityWatch/activitywatch/blob/master/LICENSE.txt",
      "requirements": "",
      "evidenceLimits": [
        "Source license declaration plus implemented app/package and documented access route checked; runtime not tested.",
        "Source review only; software was not installed or run, and individual accessibility fit was not tested."
      ],
      "reviewedAt": "2026-10-13"
    },
    {
      "id": "github:google/adk-kotlin",
      "name": "adk-kotlin",
      "purpose": "",
      "category": "Build & translate",
      "type": "Developer component",
      "platforms": [
        "Kotlin",
        "JVM"
      ],
      "sourceURL": "https://github.com/google/adk-kotlin",
      "getSoftwareURL": "https://github.com/google/adk-kotlin/blob/main/README.md",
      "licenseDeclaration": "Apache-2.0",
      "licenseURL": "https://github.com/google/adk-kotlin/blob/main/LICENSE",
      "requirements": "Kotlin/JVM app development",
      "evidenceLimits": [
        "Apache-2.0 agent toolkit; Kotlin development and model-provider setup required.",
        "Source review only; software was not installed or run, and individual accessibility fit was not tested."
      ],
      "reviewedAt": "2026-10-13"
    }
  ]
}
```

**Filtered Example Request:**

```bash
curl -X GET "https://prod.api.market/api/v1/myworkingmemory/software-library-search/search?query=notes&platform=macOS&limit=2" \
  -H "x-api-market-key: YOUR_API_KEY"
```

**Example Response (200 OK):**

```json
{
  "version": "c8f943dfc40e07f096a1fc0ae00040dc14d1df32657e8ae91f82d01aed3fab9b",
  "total": 1,
  "records": [
    {
      "id": "github:laurent22/joplin",
      "name": "Joplin",
      "purpose": "Keeps searchable notes, attachments and clipped web pages together so information can be found without remembering its location.",
      "category": "Remember & organize",
      "type": "app",
      "platforms": [
        "macOS",
        "Windows",
        "Linux",
        "Android",
        "iOS"
      ],
      "sourceURL": "https://github.com/laurent22/joplin",
      "getSoftwareURL": "https://joplinapp.org/help/install/",
      "licenseDeclaration": "AGPL-3.0-or-later",
      "licenseURL": "https://github.com/laurent22/joplin/blob/dev/LICENSE",
      "requirements": "Capture and organization still require a user workflow; retrieval is not evidence of treating memory impairment.",
      "evidenceLimits": [
        "Source license declaration plus implemented app/package and documented access route checked; runtime not tested.",
        "Source review only; software was not installed or run, and individual accessibility fit was not tested."
      ],
      "reviewedAt": "2026-10-13"
    }
  ]
}
```

#### 2. Get a record

Returns one record by its `id`, with the same fields as a search result. Pass the exact `id` returned by `/search` in the `id` query parameter, URL-encoded: `github:laurent22/joplin` becomes `github%3Alaurent22%2Fjoplin`. Do not put the id in the URL path: the API only accepts it as a query parameter.

**Method:** `GET`

**Endpoint:** `/records`

**Parameters:**

| Name | In | Description | Type | Required | Default | Max Value |
|---|---|---|---|---|---|---|
| `id` | query | The exact record id returned by `/search`, URL-encoded, for example `github%3Alaurent22%2Fjoplin`. | string | Yes | - | 200 characters |

**Example Request:**

```bash
curl -X GET "https://prod.api.market/api/v1/myworkingmemory/software-library-search/records?id=github%3Alaurent22%2Fjoplin" \
  -H "x-api-market-key: YOUR_API_KEY"
```

**Example Response (200 OK):**

```json
{
  "id": "github:laurent22/joplin",
  "name": "Joplin",
  "purpose": "Keeps searchable notes, attachments and clipped web pages together so information can be found without remembering its location.",
  "category": "Remember & organize",
  "type": "app",
  "platforms": [
    "macOS",
    "Windows",
    "Linux",
    "Android",
    "iOS"
  ],
  "sourceURL": "https://github.com/laurent22/joplin",
  "getSoftwareURL": "https://joplinapp.org/help/install/",
  "licenseDeclaration": "AGPL-3.0-or-later",
  "licenseURL": "https://github.com/laurent22/joplin/blob/dev/LICENSE",
  "requirements": "Capture and organization still require a user workflow; retrieval is not evidence of treating memory impairment.",
  "evidenceLimits": [
    "Source license declaration plus implemented app/package and documented access route checked; runtime not tested.",
    "Source review only; software was not installed or run, and individual accessibility fit was not tested."
  ],
  "reviewedAt": "2026-10-13"
}
```

#### 3. Get the catalogue summary

Returns the collection `version`, the latest `reviewedAt` date, the record `count`, every `categories` value and every `platforms` value used in the collection, and a `basis` note on how records were reviewed. Use it to build category and platform pickers, and compare `version` with the value from your last call to see whether the collection has changed. The full response is about 21 KB because the platform list holds over 300 entries, many of them detailed setup descriptions rather than short platform names.

**Method:** `GET`

**Endpoint:** `/catalogue`

**Parameters:**

| Name | In | Description | Type | Required | Default | Max Value |
|---|---|---|---|---|---|---|
| - | - | This endpoint takes no parameters. | - | - | - | - |

**Example Request:**

```bash
curl -X GET "https://prod.api.market/api/v1/myworkingmemory/software-library-search/catalogue" \
  -H "x-api-market-key: YOUR_API_KEY"
```

**Example Response (200 OK, abridged):**

```json
{
  "version": "c8f943dfc40e07f096a1fc0ae00040dc14d1df32657e8ae91f82d01aed3fab9b",
  "reviewedAt": "2026-10-19",
  "count": 227,
  "categories": [
    "Accessible games",
    "Agent research",
    "Agent work",
    "agents",
    "AI evaluation"
  ],
  "platforms": [
    "3D printer or embosser for tactile output",
    "Already configured local model aliases; generated code executes in a fresh temporary Git worktree",
    "Alternative Python 3 CLI and Tkinter GUI routes",
    "Android"
  ],
  "basis": "Individually reviewed public software sources and documented software routes. Each record retains its own source-review date. Software was not installed or run; accessibility and individual fit were not tested."
}
```

The full response lists 60 categories and 320 platform values.

## Status Codes

| Code | Status | Description |
|---|---|---|
| 200 | OK | The request succeeded. Search responses carry `version`, `total` and `records`; `/records` returns the record object itself. |
| 400 | Bad Request | A parameter is out of range or too long, for example `limit` above 25, `offset` above 10,000, a `query` over 200 characters, or a missing `id` on `/records`. |
| 401 | Unauthorized | The `x-api-market-key` value is invalid or expired. |
| 403 | Forbidden | Your subscription does not include this endpoint. |
| 404 | Not Found | No record has the `id` you passed to `/records`, or the path is wrong (for example the id was put in the URL path instead of the query). |
| 429 | Too Many Requests | You went over 10 requests per second, or you have used all the calls in your plan. |
| 500 | Internal Server Error | Something failed while processing the request. Retry after a short delay. |
| 503 | Service Unavailable | The service is temporarily unavailable. Retry after a short delay, honouring any `Retry-After` header. |

## Rate Limits

The Free and Pro plans each allow 10 requests per second. The Free plan is a one-time 7-day trial with 50 calls, and the Pro plan includes 1,000 calls per month. Both plans stop at their allowance: there are no overage calls, so once the calls are used up further requests return `429 Too Many Requests` until the Pro allowance renews or you move from the Free trial to Pro. Every call counts as one call whichever endpoint you use, so fetch 25 records per search page instead of 10 to make the most of your allowance. Requests above the per-second limit also return `429`; back off before retrying and honour any `Retry-After` header. In our tests the endpoints answered in about 0.7 to 2.7 seconds.

## FAQ

**How do I read the whole collection without searching?**

Call `/search` with no `query`, `limit=25` and `offset=0`, then raise `offset` by 25 on each call until you have read `total` records. When the response `version` changes between pages, the collection was updated while you were paging, so start again from `offset=0` if you need a consistent copy.

**Why does `/records` return an error when I put the id in the path?**

Record ids contain a colon and a slash, such as `github:laurent22/joplin`, so they cannot go in the URL path. Pass the id as a query parameter and URL-encode it: `/records?id=github%3Alaurent22%2Fjoplin`.

**Does the API run or test the tools and models it lists?**

No. It returns descriptions, source links, setup links, licenses and requirements. The `evidenceLimits` field on each record states what was checked, and most records say the software was reviewed from its source and was not installed or run.

**Which values can I use for `category` and `platform`?**

Call `/catalogue` once and use the values in its `categories` and `platforms` lists. Some categories have close variants, such as `Audio & recordings` and `Audio and recordings`, so offer the exact values from the catalogue rather than typing your own.

**How often is the collection updated?**

The provider adds and reviews finds throughout the day. Each record has its own `reviewedAt` date, and `/catalogue` returns the latest review date and a `version` value you can compare with the one from your last call.

**Why are some fields empty?**

Some records have an empty `purpose` or `requirements` string when the reviewer did not add one. Treat an empty string as "not provided" and fall back to the `name`, `type` and `evidenceLimits` fields.

**Is the Free plan renewable?**

No. The Free plan is a one-time trial of 50 calls over 7 days. After that, the Pro plan gives 1,000 calls per month at 10 requests per second.

## Contact Us

If you have any questions or need assistance, you can reach out to us through the embedded chat on the bottom right corner of the page or send an email to hello@magicapi.com.


GitHub tools API for software discovery: search curated open source tools and Hugging Face models with source links, licenses and setup requirements.