Search the web, fetch pages, run research and batch tasks through Linkup's AI web-search API.
Linkup ships in the w6w first-party pack. It declares 18 actions, 3 health checks, and the host runs its code in a sandbox that never sees the credential.
io.w6w.linkupLinkup is a web search and fetch API built for LLMs and agents. This app exposes it to a workflow, authenticated with an API key.
A workflow can search the web at four depths (flash, fast, standard, deep) and get back ranked sources with content snippets, a sourced natural-language answer (optionally with inline citations), or JSON matching a schema it supplies, with or without the sources behind it. It can fetch a single page as clean markdown (optionally rendering JavaScript, using the pro retrieval mode, pulling images or raw content, or extracting typed JSON with a schema), run a multi-step research agent, batch up to 100 search, fetch and research jobs asynchronously, and follow any of those jobs by id. It can also search and fetch the web as it was on a past day (Rewind, beta), run the closed-beta Extract agent that turns a listing page into rows, call the OpenAI-compatible responses endpoint, and read the account’s credit balance.
Every call is billed in credits. Search, fetch and research are synchronous or asynchronous depending on the action: research, extract and tasks return an id to poll.
Three routes to the same 18 actions. The Workflow tab is generated from Linkup's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
create-response Linkup's proxy of the OpenAI Responses API: a web-grounded answer from `linkup-standard` (faster) or `linkup-deep` (more thorough).
extract-create Start an asynchronous Extract task that turns a page into rows of structured data, delivered as a downloadable NDJSON file. Closed beta: the organisation must be enabled (403 otherwise). Poll with Get Extract Task.
extract-get Read one Extract task. When completed, `output.resultUrl` downloads the rows as NDJSON for 24 hours.
fetch Fetch one web page (HTML up to 20 MB or PDF up to 100 MB) as clean markdown, optionally with typed JSON extracted by a schema.
research-create Start an asynchronous research agent that navigates the web to produce a sourced answer or structured JSON (typically 2 to 20 minutes). Poll with Get Research Task.
research-get Read one research task. Once completed, `output` holds the answer and sources.
rewind-fetch Return a page as it was by the end of a given UTC day. The URL must match the stored URL exactly: Linkup applies no normalisation (404 if it is not stored).
rewind-search Search the web as it was by the end of a given UTC day.
search Search the web with Linkup at a chosen depth (flash, fast, standard, deep) and return ranked sources, a sourced answer or schema-shaped JSON, depending on the output type.
search-answer Search the web and return a natural-language answer with the sources behind it, optionally with inline citations.
search-results Search the web and return ranked sources with content snippets, ready to ground an LLM.
search-structured Search the web and return JSON matching a JSON Schema you supply, optionally with the sources used.
tasks-create Submit a batch of 1 to 100 asynchronous search, fetch, research or extract tasks in one call. Each is { type, input } with the same input as the synchronous endpoint. Poll with Get Task.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying Linkup's real ids.
{
"manifestVersion": "2",
"name": "linkup-example",
"steps": [
{
"id": "create-response",
"uses": {
"app": "io.w6w.linkup",
"action": "create-response",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"input": "<input>",
"model": "<model>"
}
}
]
}create-response credits-balance extract-create extract-get extract-list +13 more actions available
Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for Linkup, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to Linkup. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "create-response",
payload: {
input: "<value>",
model: "<value>",
// instructions: "<value>",
// format: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value); npm install -g @w6w/cli w6w run conn_YOUR_CONNECTION_ID --action create-response --payload '{"input":"<value>","model":"<value>"}' Give an AI agent Linkup — without giving it Linkup's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.linkup#create-response",
"input": {
"input": "<input>",
"model": "<model>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's Linkup connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
Linkup's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. Linkup itself is MIT, and the runtime that executes it is source-available (FSL).
Linkup declares its own checks, so its health is a property of the app rather than something the host guesses at.
The Search API and Fetch API monitors on status.linkup.so (OpenStatus). The dashboard and MCP server monitors are shown but do not drive the verdict.
Unauthenticated GET /v1/credits/balance. Linkup's JSON `UNAUTHORIZED` refusal passes: it proves the application is answering.
Credits remaining, from GET /v1/credits/balance. Free to call.