AgentLink: On-demand AI agents you can assemble into teams

4 points by nraychaudhuri 4 days ago

Over the past few weeks, I’ve been building AgentLink — a platform where you can discover and deploy AI agents trained for specific tasks like code review, market research, or sales outreach.

The idea is to make it feel like hiring a freelance team — but powered entirely by AI.

Here’s what AgentLink enables:

- Find agents trained for narrow tasks (e.g., code reviewers, data summarizers, lead generators) - Assemble teams instantly, like snapping together Lego blocks - Hire agents on-demand and scale usage up or down as needed - Track agent output, provide feedback, and iterate quickly.

It’s built on a custom system that handles agent orchestration, memory, and inter-agent communication.

Still very early — I’m currently testing with solo developers and small teams.

I’d love feedback from anyone who’s worked with AI agents, multi-agent frameworks, or automation tooling.

What technical (or product) challenges would you anticipate in scaling a system like this?

noizethievery 3 days ago

Hi, I can solve this - I've built similar solutions for YC startups. *What I'll deliver:* 5-email welcome sequence Automated follow-up campaigns Segmentation strategy *Timeline:* 24 hours *Investment:* £187 Happy to share working code examples if helpful. Quick question: What's your target timeline for this?

Best, Alex Thomson Senior Technical Consultant

manishamahawar 4 days ago

Hey — just watched this video and thought it was pretty relevant to this : https://youtu.be/L5jhEYofpaQ?si=Q8qqYisNi1AWZMi8

He makes a strong case that agents are the next big thing — and that this space is right at the frontier.

Feels like we’ll all be building our own agents soon and your AgentLink could become a crucial tool in that shift.

  • nraychaudhuri 4 days ago

    Thanks. Here is the loom demo video of the product:

    https://www.loom.com/share/e0fb8f5e7b3b44acb84b5de27e16388f?...

    Love to get your feedback.

    • kingkongjaffa 4 days ago

      No offense but do you have any better use cases?

      Automating posting AI written linkedin posts is basically littering the internet with trash as a service.

      • nraychaudhuri 2 days ago

        No offense taken—LinkedIn-post bots aren’t the point. Think UNIX pipes for business ops: small, single-skill agents you chain, swap, and reuse.

        Here are some concrete examples:

        - Financial KPI radar pull-stripe -> calc-kpis -> write-digest — 2-hour spreadsheet grind becomes a cron job.

        - Support triage classify-ticket -> fetch-answer -> route-edge — weekend inbox time drops 4 h → 30 m.

        - Bug-fix conveyor parse-logs -> rank-severity -> draft-patch — Great for tiny teams with no dedicated QA.

        - Receipts → Books ocr-receipt -> map-account -> push-p&l — zero-touch bookkeeping until tax season.

        I'm still tuning the granularity—“one verb per agent” feels right so far. Thoughts?

      • babyent 3 days ago

        Anything more is hard. So it will need real engineers and visionaries.