MANAGED WEB API
Search gives your model sources. It still has to build the answer.
Tabstack reads the sources, checks what is missing, reconciles conflicting information, and returns a cited answer in one call. The same API also returns clean text, matching JSON, or a completed public web task.
10,000 FREE CREDITS · NO CREDIT CARD REQUIRED · NEVER TRAINED ON
The latest Node.js LTS release, Node.js 24, which is currently the Active LTS line until its End-of-Life on April 30, 2028, introduces several changes impacting production deployments.
HeroDevs Blog | Node.js Version Support: EOL Dates and Latest Releases (July 2026)
herodevs.comNode.js 24 LTS: Upgrade from Node 22 in 2026
pkgpulse.comRecorded run · September 22, 2026
Search planning, source retrieval, page reading, reconciliation, and citations happened inside this call.
Your own input runs in the console playground. Sign in or create a free account with 10,000 credits.
THE PROBLEM AFTER SEARCH
Sources are not the answer.
1call you make
Tabstack completes that answer-building workflow inside the call.
A search API can return relevant links and snippets. Your model still has to choose and read sources, identify gaps, search again, reconcile disagreement, synthesize the result, and attach citations.
Read the research guide- Query planning
- Retrieval
- Gap detection
- Re-querying
- Reconciliation
- Claim-level citation
THE PRODUCT
Ask for the output you need.
Question in. Cited answer from live sources out.
Tabstack plans the search, reads the sources, fills gaps, synthesizes the answer, and attaches citations before the response reaches your model or application.
DocsURL in. Clean text or matching JSON out.
Pass a page and the shape your application needs. Fetching, rendering, content extraction, and schema enforcement happen before the response comes back.
DocsPublic website and task in. Completed result out.
Describe the task and the starting URL. Tabstack handles navigation and interaction inside the call and returns the result with task events. The hosted Automate endpoint is limited to public sites and cannot log in; Pilo is the self-hosted path.
Docsplan
read 7 pages
cite
answer
two runtimes shipped it
ollama.com/blog
docs.openwebui.comjson
{ "price": 180.00 }markdown
# Pegasus Premiumchunks
navigate
fill
confirm
a cited answer
two runtimes shipped it
ollama.com/blogclean markdown
# Pegasus Premiummatching json
{ "price": 180.00 }a completed task
fare found · reportedWHERE TO START
Start with a job, not a tool chain.
- Get a current answer with citationsResearch
- Read a link or current documentationExtract
- Keep a source set or knowledge base currentExtract
- Turn a page into a typed objectExtract
- Complete a task on a public websiteAutomate
TRUST
How your data is handled.
Never trained on. Private by default. Built by Mozilla. The Trust page explains what is processed, the account controls, and where self-hosted Pilo fits.
Not trained on your payloads
Tabstack does not train models on customer payloads. Detailed data collection is off unless the account opts in.
How data is handledWhat hosted Automate sends
The hosted Automate endpoint sends task text and page content to a contracted model provider. The Privacy Notice lists the contracted services.
Privacy NoticeOpen source underneath
Pilo is the open-source automation engine and the self-hosted path when you need control of the browser.
Pilo on GitHubWhen we are the wrong choice
- Your hosted assistant already handles the job well enough.
- You need an air-gapped system, or no hosted processing at all.
- You need automation behind a login.
- The web layer is core to your product and you want to own the loop.
INTEGRATIONS
Put the result inside the stack you already run.
Install Tabstack where your model or application already works. Start with a Research call, then add extraction or public-site automation when the workflow needs it.
Python SDK
- Install
- pip install tabstack
- First call
- client.agent.research(query='...')
CLI
- Install
- curl -fsSL https://tabstack.ai
/install.sh | sh - First call
- tabstack agent research "..."
TypeScript SDK
- Install
- npm i @tabstack/sdk
- First call
- client.agent.research({ query: '...' })
HTTP
- Install
- no install
- First call
- POST /research
COMMON QUESTIONS
Common questions
Do you train on what I send?
No. Tabstack does not train models on customer payloads, and detailed data collection is off unless your account opts in.
Can I run this against a model I control?
Yes. Tabstack is built for people and teams running their own models, whether that is on local hardware, in your own cloud account, or through a provider while you keep the application layer.
What does each endpoint receive?
The Trust page explains what is processed, what is never trained on, and where hosted processing ends and self-hosted Pilo begins.
Should I just build the loop myself?
If the web layer is core to your product, build it. Call Tabstack when the result is what matters and you would rather not maintain search, fetch, parsing and citation logic.
How is this different from a search API?
A search API returns links. Tabstack returns the finished answer with its sources attached, so your model gets the answer without reading every page itself.