Latest
Give a LlamaIndex.TS agent live web access
A hands-on walkthrough: add five live web tools to a LlamaIndex.TS agent alongside whatever you have indexed, read pages, pull typed fields, and research questions across sources on demand.
Read MoreTabstack Blog
Blog
Add web tools to an OpenAI Agents SDK agent with Tabstack
A hands-on walkthrough: hand Tabstack's tools to an OpenAI Agents SDK agent, and understand why the adapter registers them with strict mode off without losing input validation.
·7 min read
Give a Claude Agent SDK app web access
A hands-on walkthrough: wire Tabstack's ready-built in-process MCP server into a query() call, understand the mcp__tabstack__ namespacing, and give Claude web access with no separate server to run.
·8 min read
Build a web-connected Mastra agent with Tabstack
A hands-on walkthrough: spread Tabstack's tools into a Mastra Agent for schema-enforced extraction, research, and browser automation, one import, plus the production knobs and trimming decisions that follow.
·7 min read
Give an eve agent web access with Tabstack
A hands-on walkthrough: eve reads its tools from a directory, so wiring in Tabstack is five one-line files, schema-enforced extraction, research, and browser automation, with no array to assemble.
·7 min read
Stream a research assistant with the Vercel AI SDK and Tabstack
A hands-on walkthrough: give the model Tabstack's web tools with the Vercel AI SDK, then stream a cited answer token by token so users see progress forming instead of staring at a spinner.
·7 min read
Build a competitor brief agent with LangChain.js and Tabstack
A hands-on TypeScript walkthrough: pull pricing off a page, research positioning across sources, generate the judgement calls, then assemble a LangChain.js agent that produces a competitor brief with sources attached.
·9 min read
Build a research agent that cites its sources with LangChain and Tabstack
A hands-on Python walkthrough: call a Tabstack tool on its own, extract typed fields from a page, then hand all five tools to a LangChain agent that answers open questions with citations, in roughly forty lines with no browser to install.
·8 min read
Turn any page into clean markdown or typed JSON, cheaply
Research and browser agents get the demos. But most of the time the answer is just "fetch this page and give me the clean version," which is a cheaper, deterministic call, not an agent. Turn any URL into clean markdown, or into the exact typed fields you asked for. Here's why it's the right default ...
·6 min read
Tabstack on Stripe Projects
AI agents cannot manually navigate web dashboards to provision API keys. Tabstack is now integrated with Stripe Projects, allowing developers and autonomous software to provision browsing infrastructure, manage account lifecycles, and handle usage-based billing entirely through the Stripe CLI with a...
·4 min read
Introducing Schema Source: Programmatic Blueprinting for Web Data Extraction
Building schemas for AI web scraping is tedious. We built Schema Source, an independent API that instantly reverse-engineers any webpage to generate a production-ready JSON schema. Learn how to combine it with Tabstack to automate programmatic blueprinting, extraction, and data transformations with ...
·4 min read
Hardening AI Web Agents: How We're Securing Tabstack Against Indirect Prompt Injection
Following a disclosure from Brave, we patched an indirect prompt injection vulnerability in our automation endpoint. This report details the original exploit behavior and the new isolation boundaries added to the Mozilla Pilo engine.
·7 min read
Upgrading Pilo to Support Human-in-the-Loop Browser Automation
What happens when an AI web agent needs context it does not have? With Pilo Interactive Mode, it no longer has to guess or fail. This new beta feature lets Pilo pause its reasoning loop, ask the user or a parent agent for the missing data, and seamlessly complete the task.
·5 min read
Tabstack's New Pricing Plans
Tabstack's pricing is evolving to give you more control and flexibility. Enjoy clearer, consistent billing with our new free 10k credit Trial, Pay As You Go Individual plan, or predictable monthly subscriptions. Build without interruption and scale at your own pace!
·6 min read
Not Every Page Needs a Browser
We rearchitected how Tabstack fetches pages to be faster by default, and added a new effort parameter that puts you in control. Start lightweight, escalate to full browser rendering only when you need it, and build adaptive pipelines that retry intelligently.
·4 min read
Scaling Web Automation with Pilo and Agentic AI
Building reliable AI web agents is incredibly complex. To solve this, we are open sourcing Pilo. Pilo is a robust execution engine that uses the accessibility tree, smart context compression, and agentic reasoning loops to reliably navigate the modern web. Run it entirely on your own hardware.
·6 min read
Tabstack Research: Verified Answers from the Open Web
Tabstack Research moves the autonomous reasoning loop into the infrastructure layer. We handle the discovery, extraction, and verification required to bridge the synthesis gap. Access high-fidelity, structured data with inline citations instead of raw HTML noise or model hallucinations.
·11 min read
Tabstack: Browsing Infrastructure for AI Agents
Building browser infrastructure is a hidden tax on AI development. Tabstack is the developer API that turns the chaotic web into clean inputs for your agents. From rendering SPAs to complex automation, we handle the orchestration so you can focus on reasoning. Reliable, scalable, and built by Mozill...
·6 min read
Add web tools to an OpenAI Agents SDK agent with Tabstack
A hands-on walkthrough: hand Tabstack's tools to an OpenAI Agents SDK agent, and understand why the adapter registers them with strict mode off without losing input validation.
·7 min read
Give a Claude Agent SDK app web access
A hands-on walkthrough: wire Tabstack's ready-built in-process MCP server into a query() call, understand the mcp__tabstack__ namespacing, and give Claude web access with no separate server to run.
·8 min read
Build a web-connected Mastra agent with Tabstack
A hands-on walkthrough: spread Tabstack's tools into a Mastra Agent for schema-enforced extraction, research, and browser automation, one import, plus the production knobs and trimming decisions that follow.
·7 min read
Give an eve agent web access with Tabstack
A hands-on walkthrough: eve reads its tools from a directory, so wiring in Tabstack is five one-line files, schema-enforced extraction, research, and browser automation, with no array to assemble.
·7 min read
Stream a research assistant with the Vercel AI SDK and Tabstack
A hands-on walkthrough: give the model Tabstack's web tools with the Vercel AI SDK, then stream a cited answer token by token so users see progress forming instead of staring at a spinner.
·7 min read
Build a competitor brief agent with LangChain.js and Tabstack
A hands-on TypeScript walkthrough: pull pricing off a page, research positioning across sources, generate the judgement calls, then assemble a LangChain.js agent that produces a competitor brief with sources attached.
·9 min read
Build a research agent that cites its sources with LangChain and Tabstack
A hands-on Python walkthrough: call a Tabstack tool on its own, extract typed fields from a page, then hand all five tools to a LangChain agent that answers open questions with citations, in roughly forty lines with no browser to install.
·8 min read
Turn any page into clean markdown or typed JSON, cheaply
Research and browser agents get the demos. But most of the time the answer is just "fetch this page and give me the clean version," which is a cheaper, deterministic call, not an agent. Turn any URL into clean markdown, or into the exact typed fields you asked for. Here's why it's the right default ...
·6 min read
Tabstack on Stripe Projects
AI agents cannot manually navigate web dashboards to provision API keys. Tabstack is now integrated with Stripe Projects, allowing developers and autonomous software to provision browsing infrastructure, manage account lifecycles, and handle usage-based billing entirely through the Stripe CLI with a...
·4 min read
Introducing Schema Source: Programmatic Blueprinting for Web Data Extraction
Building schemas for AI web scraping is tedious. We built Schema Source, an independent API that instantly reverse-engineers any webpage to generate a production-ready JSON schema. Learn how to combine it with Tabstack to automate programmatic blueprinting, extraction, and data transformations with ...
·4 min read
Hardening AI Web Agents: How We're Securing Tabstack Against Indirect Prompt Injection
Following a disclosure from Brave, we patched an indirect prompt injection vulnerability in our automation endpoint. This report details the original exploit behavior and the new isolation boundaries added to the Mozilla Pilo engine.
·7 min read
Upgrading Pilo to Support Human-in-the-Loop Browser Automation
What happens when an AI web agent needs context it does not have? With Pilo Interactive Mode, it no longer has to guess or fail. This new beta feature lets Pilo pause its reasoning loop, ask the user or a parent agent for the missing data, and seamlessly complete the task.
·5 min read
Tabstack's New Pricing Plans
Tabstack's pricing is evolving to give you more control and flexibility. Enjoy clearer, consistent billing with our new free 10k credit Trial, Pay As You Go Individual plan, or predictable monthly subscriptions. Build without interruption and scale at your own pace!
·6 min read
Not Every Page Needs a Browser
We rearchitected how Tabstack fetches pages to be faster by default, and added a new effort parameter that puts you in control. Start lightweight, escalate to full browser rendering only when you need it, and build adaptive pipelines that retry intelligently.
·4 min read
Scaling Web Automation with Pilo and Agentic AI
Building reliable AI web agents is incredibly complex. To solve this, we are open sourcing Pilo. Pilo is a robust execution engine that uses the accessibility tree, smart context compression, and agentic reasoning loops to reliably navigate the modern web. Run it entirely on your own hardware.
·6 min read
Tabstack Research: Verified Answers from the Open Web
Tabstack Research moves the autonomous reasoning loop into the infrastructure layer. We handle the discovery, extraction, and verification required to bridge the synthesis gap. Access high-fidelity, structured data with inline citations instead of raw HTML noise or model hallucinations.
·11 min read
Tabstack: Browsing Infrastructure for AI Agents
Building browser infrastructure is a hidden tax on AI development. Tabstack is the developer API that turns the chaotic web into clean inputs for your agents. From rendering SPAs to complex automation, we handle the orchestration so you can focus on reasoning. Reliable, scalable, and built by Mozill...
·6 min read
Start automating the web in minutes.
The model, the browser, and the orchestration all run on Tabstack. You just make the call.