
Agentic Scout
io.github.pooriaarabPublished by GitHub account pooriaarabv0.1.0Updated Oct 27, 2026
Find relevant threads and viral content ideas across social platforms.
The publisher's own server: agentic.no
Overview
Searches Reddit, Hacker News and RSS for public threads, then scores, ranks and turns them into template-based content idea cards.
- What you can do
- Search public threads on Reddit, Hacker News and RSS feeds, or score supplied fixture records
- Score and rank threads with local heuristics for relevance, engagement, buying-intent terms and competition
- Generate up to ten template-based content idea cards with source links and disclosure text
- Summarize text and run local dispatch that produces receipts without external delivery
- Report process-local dispatch units as a cost report
- When to use it
- Useful for developers researching public conversations who want on-demand thread search and local ranking inside their own process, for example to feed a research dashboard or to shortlist threads before drafting content ideas. It is not a monitoring or alerting service.
- Requirements
- Runs locally as an npm package via npx and needs Node.js 18 or later. No credentials, environment variables or .env loader are required. Network access to Reddit, Hacker News (Algolia) and RSS endpoints is needed for live searches; fixtures work offline.
Installation
In SourceWeft
- Open Agentic Scout in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
agentic-scout
[npm version] [Monthly downloads] [MIT license]
agentic-scout is an open-source thread discovery library for developers who research public conversations. It searches Reddit, Hacker News, and RSS feeds. Local scoring and fixed templates produce ranked threads and draft content ideas.
Install
Requires Node.js 18 or later. Includes TypeScript declarations and two direct runtime dependencies.
Find your first thread
Save this as scout-demo.mjs, then run node scout-demo.mjs. No credentials, environment variables, or .env loader are required.
Success prints 1 Why Node.js tools is getting attention right now and an empty cost report.
Fixtures run locally. Calling searchThreads without source arguments immediately queries Reddit and Hacker News.
Preview a network search with the CLI:
This prints dryRun: true and approvalRequired: true. Add --live after approving network access yourself.
How it works
Why it exists
Developers can combine public thread retrieval and local scoring in their own applications. F5Bot provides keyword alerts; Feedly provides source-following workflows. This package exposes programmable search results and draft ideas.
Compare
Scope checked on 2026-10-14 against linked official documentation.
Coverage
The dispatcher implements four source paths: three network sources and one fixture path. No credential-checked integrations are claimed. Network availability depends on the upstream endpoint.
Common use cases
- Node.js developers retrieve Hacker News stories for a research dashboard.
- TypeScript teams rank supplied Reddit records with local heuristics.
- MCP client users score threads before drafting content ideas.
- RSS tool builders filter feed entries using a query in a Node.js script.
Features
- Search: SDK options include a subreddit, RSS URL, result limit, fixtures, and injected fetch.
- Scoring: Scores cover title-based relevance, engagement velocity, buying-intent terms, and competition.
- Ranking: Total score orders supplied records; titles break ties.
- Ideas: Up to ten template-based cards include source links and disclosure text.
- Utilities: Shared SDK helpers provide text statistics, guards, local routing, and HTTP routing.
Runtime surfaces
The Node.js SDK exports searchThreads, scoreOpportunity, rankThreads, and suggestIdeas, plus shared core utilities.
Seven CLI commands: search, trends, ideas, summary, dispatch, costs, and mcp.
Use --help or -h for help; --version or -V prints the version.
Search and ideas require --query; trends uses the fixed query trending.
Search commands accept --platforms, --dry-run, and --live.
Summary accepts --text and --top; dispatch accepts --text and --dry-run.
Seven MCP tools: summarize_text, dispatch_text, score_opportunity, rank_threads, suggest_ideas, search_threads, and get_cost_report.
MCP network search requires both live: true and approved: true.
A repository Worker handler provides GET /health and POST routes /summarize, /score, /rank, and /ideas.
Users deploy that source themselves. It is not a managed endpoint or package export.
Delivery and behavior limits
Search has no pagination, retries, built-in timeout, cache, or deduplication. Sources run sequentially; a failure stops the whole search. Reddit and Hacker News requests each ask for 25 records. The combined default limit is 50. RSS uses a basic regular-expression parser; it does not fully implement RSS or Atom.
Relevance uses title length and tags, rather than similarity to the search query. Scores do not establish buying intent or predict virality. Idea generation uses fixed templates without a model API. LinkedIn and X (Twitter) are idea labels, not publishing integrations.
The CLI accepts RSS but cannot supply rssUrl; use the SDK or MCP for feeds.
The CLI accepts --file, but its default reader is absent. Use --text for statistics and local dispatch.
CLI live search needs --live; there is no separate approval flag. SDK searches perform network calls directly.
Local dispatch produces receipts without external delivery. Generic SDK HTTP routing has a ten-second default timeout and no retries. Cost reports store process-local dispatch units, not billing totals. Searches do not record costs.
Integration guides
Use the fixture quickstart for offline applications. For RSS, pass sources: ["rss"] and rssUrl to the MCP search tool.
In the SDK, use searchThreads(query, ["rss"], { rssUrl }).
No public documentation site is configured in the repository metadata.
Status
Version 0.1.0 is an early release; interfaces can change. Report issues through GitHub support. Public case studies are not yet published in this repository. Review source facts and draft disclosures before publishing content.
FAQ
What is agentic-scout?
agentic-scout is a public thread search and ranking library. It also creates template-based content idea cards.
Which integrations are verified?
agentic-scout implements Reddit, Hacker News via Algolia, RSS, and fixtures. The local fixture example is verified; live endpoint access is pending.
Does agentic-scout provide a managed monitoring service?
agentic-scout runs in your own process. It does not provide scheduled monitoring or managed alerts.
How is agentic-scout different from F5Bot?
agentic-scout exposes on-demand search and local ranking. F5Bot monitors keywords and emails alerts.
How is agentic-scout different from Feedly?
agentic-scout returns thread records through a developer API. Feedly offers source-following workflows.
Does it guarantee viral ideas?
agentic-scout makes no virality prediction. Its scores are heuristics and its ideas are fixed templates.
Does search run in dry-run mode?
agentic-scout CLI search previews by default, while SDK search runs directly. MCP search requires explicit live and approval inputs.
Does it publish to LinkedIn or X?
agentic-scout uses LinkedIn and X as idea labels. It has no social publishing adapters.
License
MIT.
Source: README.md at commit 270dedb
Tools
0Version history
1- v0.1.0LatestOct 15, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Other Agentic integrations
agentic.no
no.agenticOfficial publisher
A remote MCP endpoint that helps an assistant find GPU sellers for AI inference by model, ordered cheapest first, with per-call payment via x402.
Agentic Engineering Periodic System
academy.aseOfficial publisher
Agentic Engineering Periodic System's hosted MCP server, published by ase.academy; no sign-in required; Streamable HTTP; 4 tools.
Agentic RL: Credit Assignment and CLI Agents
space.hf.hoyant-su-agentic-rlOfficial publisher
Lets an assistant filter agentic reinforcement-learning methods by supervision, critic and task setting, and retrieve source links and BibTeX.

Agentic Video
Published by GitHub account tlq5l
Local stdio MCP server that lets a text-and-image assistant inspect video files: contact sheets, transcripts, cuts, loudness and sync checks.

Agentic Os
Published by GitHub account kinnister
Read-only MCP server that exposes the agentic-os governance, SDLC and quality-engineering methodology to an assistant, without executing anything.