curated-finds

Search Myworkingmemory Curated Finds software and models through Apify, RapidAPI or Direct.

Package Information

Released: this month
Downloads: 0 weekly / 0 monthly
Latest Version: 0.3.0
Author: Myworkingmemory LLC

Documentation

Myworkingmemory Curated Finds for n8n

Search Myworkingmemory's Software Library Search / Curated Finds catalogue from an n8n workflow or AI agent. Results contain source links, descriptions, setup requirements, license declarations, evidence limits and original review dates.

Use your existing RapidAPI subscription or Curated Finds Direct subscription. They are separate billing routes for the same catalogue; their keys are not interchangeable. Apify has a separate connection in version 0.3.0; use an Apify API token for it. MCPize and API.market keys do not work with these connections.

Apify connection — 0.3.0

Version 0.3.0 adds Connection: Apify for the existing Myworkingmemory Software Library Search Actor. The connection is checked offline; the validation record identifies the live verification scope.

  1. Select Apify and create a Myworkingmemory Curated Finds (Apify) API credential with your existing Apify API token; restrict credential HTTP access to api.apify.com.
  2. Saving, reopening or testing that credential sends only GET https://api.apify.com/v2/users/me. It starts no Actor and buys no search page; it also does not establish paid-search eligibility.
  3. Set a finite Maximum Run Charge (USD), default $0.01 (allowed $0.01–$1). The node sends one authenticated POST to https://api.apify.com/v2/acts/ghlTCYxrSCV2k1UzV/run-sync, with timeout=30, memory=256, restartOnError=false, maxTotalChargeUsd and outputRecordKey=OUTPUT.
  4. For the fixed sample, enable Fixed Preview, set Page Size 3, leave Query/Category/Platform blank and Offset 0. The same three sample records return each time; this is not a keyword search and has no catalog-search event charge, though platform/storage costs can apply.
  5. For a full search, leave Fixed Preview off and supply the normal filters. The current Actor charges $0.00599 per completed search page, including an empty page, and requires an eligible paid Apify account. Your Apify account, credit and charges are separate from RapidAPI and Direct; the node never buys a plan.

Each input item or AI tool invocation starts a separate Actor run with its own cap. Keep Retry On Fail off. The HTTP timeout is 60 seconds while the Actor runtime is capped at 30 seconds; startup can also take time. A disconnected/timed-out request may have started a billable run: inspect Apify Runs before trying again. The node makes no automatic retry, paging or second request to recover a run. The Actor's OUTPUT retains catalogue version, total, preview/search mode and evidence; nextOffset is derived from the returned page, and remaining account quota is never invented.

The credential-free Apify preview example is inactive and starts nothing until manually executed with your credential.

Installation

Self-hosted users can install n8n-nodes-curated-finds under Settings → Community Nodes, following n8n's installation guide. Search the node picker for Myworkingmemory, Curated Finds, Software Library Search or RapidAPI, then select Myworkingmemory Curated Finds.

This package is not yet verified by n8n. n8n Cloud installation requires separate verification; npm publication does not establish Cloud availability.

Connect RapidAPI

  1. Open Software Library Search on RapidAPI and use an application with access to that API.
  2. In the n8n node, select Connection → RapidAPI.
  3. Create a Myworkingmemory Curated Finds (RapidAPI) API credential. Paste that application's X-RapidAPI-Key from the RapidAPI playground.
  4. Select the saved credential. Your key is encrypted in n8n, never placed in workflow parameters or example JSON.

Saving or testing this credential sends one real search request (query=SpeechBrain, limit=1, offset=0). It counts toward the selected plan and may incur its charges. A free plan with one request can be consumed by this credential test, leaving no workflow request. Check current RapidAPI pricing, quota and bandwidth terms before running. The node does not buy or change a plan.

Requests use only:

GET https://software-library-search.p.rapidapi.com/search
X-RapidAPI-Host: software-library-search.p.rapidapi.com
X-RapidAPI-Key: <your encrypted RapidAPI credential>

Connect Direct

  1. Open Curated Finds Direct and sign in to your subscription, or review the subscription there.
  2. Select Show API key and copy your cf_… key into a Myworkingmemory Curated Finds API credential.
  3. Select Connection → Direct and that credential.

Direct remains the default to preserve existing workflows from versions 0.1.x. Saving or testing a Direct credential uses one included lookup, fixed to query=n8n, limit=1, offset=0, even with no matches. Direct currently costs $5.99/month for 1,000 lookups shared with the website reader, with a hard allowance cap. See the current offer and terms.

GET https://www.myworkingmemory.ai/api/software-library/direct/search
Authorization: Bearer <your encrypted Direct credential>

Search and output

Software → Search sends one request per input item. Use Execute Once if your workflow should make only one request. n8n hosting and AI-model costs are separate.

Field Behaviour
Query Optional text, maximum 200 characters; blank lists matching records.
Page Size Integer 1–25; default 10.
Category Optional existing filter value, maximum 100 characters.
Platform Optional existing filter value, maximum 100 characters.
Offset Integer 0–10,000; default 0. Use nextOffset explicitly to request another page.

Each input returns one page with version, total, records and nextOffset. RapidAPI's nextOffset is calculated from the requested offset and returned record count; its JSON does not report remaining plan quota. Direct additionally returns remaining. No RapidAPI quota is invented.

Output defaults to Raw (all record fields). Simplified returns ten useful fields. Selected Fields returns chosen fields plus ID, source URL, original review date and evidence limits. All modes retain the page envelope. Use n8n's Split Out node on records only after retaining page information you need. Compare catalogue version between pages.

The node supports one-page search through all three connections. It does not run or install the catalogue's tools, provide date filters, or implement the other RapidAPI endpoints. Source review does not prove runtime compatibility, accessibility fit or freshness beyond each record's review date.

Request limits

There is no automatic pagination or node-level retry. Each request goes to the selected fixed HTTPS host, refuses redirects and caps decoded responses at 1 MiB. Direct and RapidAPI workflow searches enforce 30 seconds; the Apify HTTP request enforces 60 seconds and its Actor runtime is capped at 30 seconds. n8n 2.45.5 controls credential-test timing and may wait up to five minutes; its authentication helper may retry a rejected authentication request once. Plan rules determine whether failed requests count. Keep Retry On Fail off and inspect usage before retrying a timeout, since the request may already have been processed.

Authentication, allowance and service failures remain errors. A successful empty page still uses a request. Every later page and every new AI-tool invocation is another request.

Examples and AI tools

  • RapidAPI workflow: Manual Trigger → Myworkingmemory Curated Finds → Review catalogue page.
  • Direct workflow: preserves the original Direct connection and allowance output.

Both examples are inactive, credential-free and have no saved execution data or retries. They search SpeechBrain, page size 3, offset 0. Each execution sends one request in addition to any credential test.

For an AI agent, add Myworkingmemory Curated Finds as a tool, choose the correct connection and your credential, and set a small fixed page size. usableAsTool: true is enabled. Configure agent iteration limits: repeated tool invocations consume separate requests. The publisher's credentials are never provided to customers.

Compatibility and development

Built with the official n8n declarative node CLI 0.55.1. Offline tests use n8n-core 2.45.4 and n8n-workflow 2.45.2 from n8n 2.45.5, exercising the native loader, parameter mapping, credential selection, routing and hooks with mocked HTTP. This does not claim a live API request, browser credential test or AI-model execution. See VALIDATION.md.

Use Node.js 24 LTS:

npm ci
npm run check
npm run scan:local
npm run dev

Publishing and support

Source: myworkingmemory/n8n-nodes-curated-finds. Releases use the tag-triggered .github/workflows/publish.yml GitHub Action with npm provenance and the authorized package-specific trusted publisher. Verification by n8n is a separate step.

Documentation: Curated Finds API. Support: dean@myworkingmemory.ai or Myworkingmemory support. License: MIT.

Discussion