LinkList — grounded sources for agents

Your agent, grounded in sources you vetted.

Add one endpoint to Claude Code, Codex or Cursor. Your agent searches the collection you curated and answers with the URL and the date each source was fetched.

Query the demo collection first Free · 100 queries/day · no card

A real session

$ claude mcp add --transport http linklist https://linklist.io/mcp
  added. 5 tools registered.
> what webhooks are available?
  search_sources → 8 citations · 278ms
  [1] developer.tryfinch.com · fetched 2026-07-28 · link_id 3015
  [2] plaid.com · fetched 2026-07-28 · link_id 3019

Protocol 2025-11-25. Five tools: three read, two write.

The same question, two agents

Ungrounded

Most payroll platforms support webhooks for payroll and employment events.

Plausible. Unattributable. Possibly wrong.

Grounded

Plaid sends webhooks to programmatically inform you about changes to Plaid Items or the status of asynchronous processes.1

1 plaid.com · fetched 2026-07-28 · link_id 3019

Checkable. You chose the source.

Works where you work

Bring your sources to your agent.

NotebookLM asks you to move your work into its app and use its model. LinkList runs underneath the agent you already have open.

NotebookLM

  • You open their app and work there.
  • 50 sources per notebook on the free tier, 300 on Pro.
  • Gemini on every tier. No model choice.

LinkList

  • You stay in Claude Code, Codex or Cursor.
  • Your agent queries the collection over MCP.
  • Any model. Any MCP client.

How it works

Where your agent's answer comes from.

Say your agent keeps inventing webhook event names that look right and are not. This is the whole fix.

  1. Step 01 · once, about a minute

    Paste one command

    Your coding agent gets one new tool. Nothing else about your setup changes.

    No new app to open. You stay in your editor.

    $ claude mcp add --transport http \
    linklist https://linklist.io/mcp
    added. 5 tools registered.

  2. Step 02 · once per topic

    Add the pages you trust

    Paste in the real documentation. LinkList reads each page and keeps the text, so your agent can quote it later.

    You decide what counts as a source. Nothing else gets in.

    https://developer.tryfinch.com/developer-resources/Webhooks
    https://plaid.com/docs/api/webhooks/
    https://docs.truv.com/docs/webhooks

  3. Step 03 · every time from now on

    Ask the way you already do

    Your agent reads those pages before it answers, instead of guessing from memory.

    You can open the page it used and check it yourself.

    > what webhooks are available?
    search_sources → 8 citations
    [1] developer.tryfinch.com · fetched 2026-07-28 · link_id 3015
    [2] plaid.com · fetched 2026-07-28 · link_id 3019

    Truv fires order-created when an order is first created, before any status transitions.1

    1  docs.truv.com · fetched 2026-07-28 · link_id 3017

Steps 1 and 2 happen once. Step 3 is every day after that.

Pricing

Free

$0
  • 3 collections, 100 sources
  • 100 queries per day

Pro

$12 per month
  • Unlimited collections and sources
  • 2,000 queries per day
  • Manual re-fetch, priority embedding

MCP starter pack

Verified setup configs for every client.

Copy-paste configuration for Claude Code, Claude Desktop and Cursor, plus the eval harness so you can measure your own retrieval.

One email with the pack. No sequence.

Questions

Which clients does this work with?
Anything that speaks MCP over streamable HTTP: Claude Code, Claude Desktop, Cursor and others. The server negotiates protocol 2025-11-25 and older revisions.
How is this different from giving my agent a web search tool?
Web search returns whatever ranks. This returns only what you added, and it tells your agent where each passage came from and when it was fetched.
Do you train on my sources?
No. Sources are fetched, chunked and embedded so they can be retrieved for your own queries. Libraries are private unless you publish them.
What happens when a source changes?
Content is hashed on fetch. Pro accounts can re-fetch on demand. Scheduled re-crawling is not shipped yet.
How do I know retrieval actually finds the right source?
Measure it. The test harness ships with the product, so you can point it at your own sources and see how often the right one comes back. We run it on ours before every release and publish what it says, including when it falls short.

Point your agent at sources you trust.

Free tier, no card. The endpoint is live now.