The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Bluesky Keyword Alerts listing page.
Public Apify Task configurations, validated n8n workflow examples, Make construction specifications, and live Make shared scenarios for monitoring Bluesky posts, RSS feeds, and TED procurement notices.
Created and maintained by Telemark Digital.
server.json metadata and copy for remote MCP directories.Each product directory contains public Apify Task inputs, a platform-neutral workflow contract, an n8n workflow, a Make implementation package, fixtures, and validation notes.
The TED package now includes thirteen live Apify Examples pages for high-intent procurement searches: AI procurement, cybersecurity, healthcare, SaaS/cloud software, construction, CPV 72 software, change notices and corrigenda, German managed detection, recent procurement JSON previews, EU software CPV previews, French medical equipment, Polish IT services, and renewable energy procurement.
The RSS package now includes eleven live Apify Examples pages: multi-feed keyword alerts, regex and exclusions, webhook-ready JSON, security advisories, vulnerability research, AI product news, research paper alerts, DevOps release notes, cloud platform updates, company news, and government cyber updates.
The Bluesky package now includes eleven live Apify Examples pages: keyword search, brand-handle mentions, hashtags, product launch keywords, developer tools launches, open-source maintainer discussion, AI policy posts, security keywords, data breach discussion, startup funding posts, and brand or competitor handle monitoring.
Live Make shared scenarios:
Live n8n workflow templates:
Live MCP and AI-agent directory surfaces:
bluesky-keyword-alerts, rss-keyword-monitor, ted-tender-monitorFoundational Codex, ChatGPT desktop, and Claude Code plugin packages are included under plugins. Each plugin exposes exactly one public Apify Actor through a product-specific Telemark MCP gateway route backed by Apify OAuth/MCP and includes a product-specific skill that keeps bounded one-off discovery separate from saved-Task monitoring.
The plugin packages passed repository-source installation plus live OAuth/MCP calls in Codex and Claude Code on 2026-07-23. These repository-source packages are separate from the official OpenAI and Anthropic review programs. See AI Client Plugin Marketplaces for the package layout, safe usage boundary, and current directory gates.
Remote MCP directory metadata for the same product routes is under mcp. Official MCP Registry, Smithery, and Glama listings are live for all three products. Smithery uses distinct free-capped product Worker URLs for cleaner tool discovery. Glama reports each connector as Healthy, and the public OAuth challenge, ownership, and server-card endpoints are live; see MCP Directory Listing Pack.
Public product icons are in assets/icons using clean external filenames. The AI-client plugin manifests omit optional icon fields for this release so the plugin packages remain text-only and easy to publish through GitHub's web editor.
Actor runs can commit dataset records before ending FAILED, TIMED-OUT, or ABORTED. For each terminal run they process, the public workflows fetch the run's defaultDatasetId when present, persist records under stable product-prefixed keys, and only then expose or report the terminal status. Failed empty datasets and runs without a dataset ID are reported safely. The terminal fixtures also prove replay idempotency, and the TED fixture confirms that title: null remains valid.
For Make, the construction specifications use Apify's maximum 1000-run task-runs page, one exact-key data-store cursor per Task, cursor-filtered reverse-page processing, first-run cursor priming, and an overflow stop whenever the fetched page does not contain the stored cursor run. Each publication template processes at most one selected Task run per scenario execution (maxRunsPerScenarioExecution = 1) so operation budgets stay bounded; if an account needs a longer outage window than 1000 * Apify Schedule interval, add paginated backfill before activation.
In each n8n export, dataset retrieval and Data Table upsert retry the same operation in place up to three times, waiting five seconds between attempts. These retries keep the original defaultDatasetId and stable delivery key and never rerun the Apify Task. If either operation still fails after its retries, perform manual recovery using the original run ID and its dataset; this package does not claim automatic exactly-once delivery.
The included n8n workflows and Make specifications intentionally make one non-paginated dataset request per Task run. Full retrieval therefore depends on the saved Task cap:
| Product | Saved Task cap | Fixed retrieval limit | Reserved control rows |
|---|---|---|---|
| Bluesky | maxPostsPerRun <= 100 | 100 | None |
| RSS | maxItemsPerRun <= 200 | 200 | None |
| TED | maxNewPerRun <= 999 | 1000 | Exactly one appended summary row (999 + 1 = 1000) |
Do not activate a workflow until the account owner has verified the matching saved Task setting. Raising a cap requires a separately designed, documented, and validated paginated workflow; it is outside this package.
These examples contain no credentials. Connect your own Apify and destination accounts after importing a workflow. Read SECURITY.md before publishing or modifying an export.
The examples are prepared and validated locally before publication. All three Make scenarios and all three n8n templates are publicly shared through the links above. The repository retains credential-free construction specifications rather than exported Make blueprint files.
See Account Gates for the GitHub publisher, n8n Creator Portal, optional n8n Cloud, and Make publication steps.