MANAGED WEB API
Give your model the web without giving up your privacy.
Let your model research live sources, extract the data it needs, and automate web tasks. Tabstack manages the browsers and web interactions outside your model, so you can build your agent without building its web stack.
Never trained on. Private by default.
How we protect your data →
10,000 FREE CREDITS · NO CREDIT CARD REQUIRED
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 WEB LAYER
Web access shouldn't become your next infrastructure project.
1API to connect
Build your agent. Not its web infrastructure.
Adding web access means more than connecting a search tool. There are pages to fetch and render, data to extract, and browser tasks to run. Tabstack manages that work outside your model, so you can focus on what your agent does, not how it gets around the web.
Read the docs- Web research
- Page fetching
- Browser rendering
- Data extraction
- Schema enforcement
- Browser interaction
THE PRODUCT
Your model makes the call. We handle the web.
Give your model answers it can cite.
Ask a question. Tabstack searches live sources, reads the pages, fills gaps, and brings the findings together with citations. Your model gets the research without running every search and reading every page itself.
DocsTurn web pages into data your code can use.
Pass a URL and, for structured data, the schema you need. Tabstack fetches and renders the page, strips out the boilerplate, and returns clean text or JSON that matches your schema, without a browser or parser for you to maintain.
DocsDescribe the task. Leave the browser to us.
Give Tabstack a starting URL and a task. It navigates and interacts with the website, then returns the result and task events. Hosted Automate cannot log in; use self-hosted Pilo for workflows that require authentication.
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 websiteAutomate
TRUST
Never trained on. Private by default.
Built by Mozilla. Here's what those commitments mean for the data you send to Tabstack.
Never trained on.
Tabstack and our contracted model providers do not use your payloads to train models under the agreements governing the service.
How we protect your dataPrivate by default.
You control request history. An organization admin can turn off detailed data collection in the console to stop storing new request payloads. Operational and security logs still apply, as described in our Privacy Notice.
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 GitHubKnow the limits.
- 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
Give the agent you already use a way onto the web.
Add Tabstack through an integration, SDK, CLI, or direct API call. Bring research, extraction, and browser automation into your existing workflow without running the browsers yourself.
- Install
- pip install tabstack
- First call
- client.agent.research(query='...')
- Install
- curl -fsSL https://tabstack.ai
/install.sh | sh - First call
- tabstack agent research "..."
- Install
- npm i @tabstack/sdk
- First call
- client.agent.research({ query: '...' })
- Install
- no install
- First call
- POST /research
Agent and framework integrations
COMMON QUESTIONS
Common questions
Do you train on what I send?
No. Tabstack does not train models on customer payloads.
Does my model need to operate a browser?
No. Your model or agent calls Tabstack. Tabstack handles the browser and web workflow and returns the result.
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.
Do I need to run the web infrastructure?
No. Tabstack manages the browsers and web interactions. If you need to run the automation engine yourself, Pilo is open source.
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.