How much does AgentQL cost?

Free trial with 300 API calls. Starter is $0 a month with 50 calls and 10 remote browser hours included, then $0.02 a call and $0.12 a browser hour. Professional is $99 a month with 10,000 calls and 500 browser hours, then $0.015 a call and $0.10 an hour. Enterprise is custom.

What are the best alternatives to AgentQL?

Based on documented capability coverage across the Agentic Index 14 point taxonomy, the closest alternatives to AgentQL among browser and computer-use agents are Autotab, Stagehand, Bright Data Agent Browser, Notte. Each is graded on the same public evidence, and no vendor pays for placement.

Does AgentQL support MCP and tool calling?

Full under the Agentic Index verification standard. AgentQL publishes public evidence of Model Context Protocol (MCP) support, tool calling, and API or SDK extensibility for connecting agents to external tools. Evidence basis: Python and JavaScript SDKs with API references, a REST API for queries and remote browser sessions, and an npm MCP server authenticated with the API key expose AgentQL to developers and agents. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Can AgentQL be self hosted or run in a private VPC?

Full under the Agentic Index verification standard. AgentQL publishes public evidence of self hosted or private VPC deployment, along with data residency options for controlling where agent data is processed and stored. Evidence basis: The Enterprise plan offers a fully managed dedicated cloud environment, and on premise deployment is available. No region list is published. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does AgentQL give you tracing and observability into agent runs?

Partial under the Agentic Index verification standard. AgentQL publishes some public evidence of logging or observability, but narrower than full tracing and auditability across agent runs. Evidence basis: Scheduled workflows show each job's status, the query used, the scraped data and an optional screenshot of the page at run time. Remote sessions stream the live page. AgentQL names no audit trail or run trace that explains why a step happen. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Back to vendors

AgentQL

Q: Who is AgentQL for?

developers, AI agent builders, data teams, QA and testing teams

Also known as: agentql.com

Visit site

Browser / computer-use agentfirst party productVerified 2026-09-27

Entry priceFree trial (300 API calls); Starter $0 a month, pay as you go at $0.02 a call; Professional $99 a month; Enterprise customFull pricing detail

A self healing query language for the web, with natural language selectors that find elements by meaning, survive layout changes, and return structured JSON or Playwright locators. SDKs, a REST API, an MCP server and framework integrations make it the semantic selector layer beneath agents.

AgentQL, from TinyFish, is a query language for the web. Developers describe elements and data in natural language style queries ({ products[] { name description price(integer) } }) and AgentQL finds them by meaning rather than DOM position, returning structured JSON shaped by the query or Playwright locators for acting on the page. Queries are self healing across layout changes and reusable across similar sites, and the same language extracts data from PDFs and images. Under the hood AgentQL routes each query across several public LLMs and a proprietary model, with the customer choosing fast or standard mode.

The suite spans Python and JavaScript SDKs, a REST API, an MCP server, integrations that expose AgentQL inside LangChain, LlamaIndex, Dify, Langflow, Google ADK and Zapier, a Debugger Chrome extension and a Playground for trying queries, hosted remote browser sessions with CDP access, stealth and the TinyFish Browser profile, and an experimental Dev Portal scheduler for recurring scraping jobs. Enterprise adds a dedicated cloud or on premise deployment. AgentQL is the semantic selector layer beneath agents and automations, not an agent platform itself. TinyFish's own web agent platform is listed separately.

Vendor details

Canonical URL

https://www.agentql.com

Funding status

A product of TinyFish, which also sells its own web agent platform (separate record, tinyfish); open source SDK tooling under the tinyfish-io GitHub organisation with a hosted query service

Company status

first party product

Use cases & customers

Primary use cases

natural language element selection for automationself-healing web scraping and data extractionagent web access via MCP and framework integrationsE2E testing with resilient selectorsdocument extraction (PDF and images)

Target customers

developersAI agent buildersdata teamsQA and testing teams

Deployment options

Hosted API and remote browser sessionsSDK in the customer's environmentEnterprise dedicated cloud or on premise

Integrations

Python and JavaScript SDKs come with first class Playwright integration, alongside a REST API for queries and remote browser sessions and an MCP server. Integrations expose AgentQL inside LangChain, LlamaIndex, Dify, Langflow, Google ADK and Zapier, and a Debugger Chrome extension and Playground are there for trying queries. Hosted remote Chrome sessions come with CDP access and stealth.

In practice

Your team needs prices from the same list of competitor pages every week. AgentQL's Dev Portal scheduler runs one query over that list of URLs at a set time, and each job shows its status, the scraped data and an optional screenshot of the page.

Your agent built in LangChain needs to read and act on live websites. AgentQL plugs in as a tool through the LangChain, LlamaIndex or MCP integrations, and its queries return Playwright locators that fill and submit forms, close modals and page through results, including on logged in pages.

Your developers are writing a new extraction query and want to see what it catches before it ships. The AgentQL Debugger Chrome extension and Playground run the query against a live page and highlight what it returns, so they can refine it first.

Sources & related URLs

Research sources

https://docs.agentql.com/concepts/main-concepts https://github.com/tinyfish-io/agentql https://docs.agentql.com/home

Agentic Index coverage score

5.5 / 14 capabilities · 39%

Integrations & Tool Calling	Not documented
The LangChain, LlamaIndex, Dify, Langflow, Google ADK, Zapier and MCP integrations run the other way, exposing AgentQL as a tool inside other agents. Logging into websites is handled as browser work. AgentQL names no connector catalog, custom tools or authenticated actions for working in outside systems. Sourcedocs.agentql.com/integrationsread 2026-09-27
Workflow Orchestration	Not documented
As a query and selector layer, AgentQL leaves multi step logic to the customer's Playwright scripts or agent framework. A scheduled workflow runs one query over a list of URLs. Sourcedocs.agentql.com/scraping/schedulingread 2026-09-27
Knowledge Grounding & RAG	Not documented
The page being queried is read at run time, and AgentQL extracts from PDFs and images per call. AgentQL names no retrieval layer over the customer's own knowledge. Sourcedocs.agentql.com/concepts/main-conceptsread 2026-09-27
Human Oversight & Guardrails	Not documented
Remote sessions offer a streaming URL for watching the page, and every action is taken by the customer's own script. AgentQL names no approval, takeover or guardrail mechanism. Sourcedocs.agentql.com/browser/remote-browserread 2026-09-27
Security, Identity & Governance	Not documented
API keys control access to AgentQL, and it names no other access model, SSO or attestation. TinyFish states ISO 27001, SSO and role based access for its own TinyFish platform, not for AgentQL. Sourceagentql.com/pricingread 2026-09-27
Observability & Auditability	Partial
Scheduled workflows show each job's status, the query used, the scraped data and an optional screenshot of the page at run time. Remote sessions stream the live page. AgentQL names no audit trail or run trace that explains why a step happened. Sourcedocs.agentql.com/scraping/schedulingread 2026-09-27
Memory & State Persistence	Not documented
Browser state persists. Cached credentials are saved as cookies and local storage with storage_state, and remote sessions stay alive for reconnect with shutdown_mode. AgentQL names no memory that queries read as context. Sourcedocs.agentql.com/logging-into-sites/caching-user-credentialsread 2026-09-27
Deployment & Data Residency	Full
The Enterprise plan offers a fully managed dedicated cloud environment, and on premise deployment is available. No region list is published. Sourceagentql.com/pricingread 2026-09-27
Prebuilt Agents, Templates & Packs	Not documented
Example scripts in Python and JavaScript and Playground query suggestions help with onboarding. The Enterprise plan offers ready to use datasets. AgentQL names no library of prebuilt agents, templates or packs. Sourcedocs.agentql.com/examplesread 2026-09-27
Triggers & Channel Coverage	Full
The Dev Portal scheduler runs scraping workflows on their own at a set time of day, week or month, with Run Now and pause controls. On demand runs come from the SDKs and REST API. The scheduler is experimental and capped at 2 workflows, 5 URLs and 10 runs per workflow. Sourcedocs.agentql.com/scraping/schedulingread 2026-09-27
Model Flexibility & Routing	Partial
Each query is routed across several public LLMs, among them GPT-4, Llama and Gemini, plus a proprietary model. AgentQL's infrastructure decides which to use by task complexity. The customer chooses only between fast and standard mode. Sourcedocs.agentql.com/concepts/under-the-hoodread 2026-09-27
APIs, SDKs & MCP Extensibility	Full
Python and JavaScript SDKs with API references, a REST API for queries and remote browser sessions, and an npm MCP server authenticated with the API key expose AgentQL to developers and agents. Sourcedocs.agentql.com/rest-api/api-referenceread 2026-09-27
Testing, Debugging & Optimization	Partial
The Debugger Chrome extension and the Playground run a query against a live page and highlight what it returns, so a developer can try and refine queries before use. AgentQL names no scored test cases, test harness or quality gate. Sourcedocs.agentql.com/debugger-extensionread 2026-09-27
Browser & Computer Use	Full
Queries return Playwright locators that fill and submit forms, close modals and page through results on any site, including logged in pages. AgentQL also runs hosted remote browser sessions with CDP access, stealth and the TinyFish Browser profile. Sourcedocs.agentql.com/browser/remote-browserread 2026-09-27

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Free trial (300 API calls); Starter $0 a month, pay as you go at $0.02 a call; Professional $99 a month; Enterprise custom

not captured

Free tier

What is public

All self serve tiers, included quotas, per call and per browser hour rates and rate limits are public; Enterprise is custom.

Billing mechanics

Monthly plans with included API calls and remote browser hours, then per call and per browser hour overage; rate limits and concurrent sessions scale by plan.

Cost watchouts

Overage is metered per API call and per remote browser hour, so heavy scraping or long sessions add up beyond the included quota.

Variable cost rationale

No billing model captured in retrieved materials.

Sales call required

Mixed (some tiers require a call)

Free / trial

Free trial with 300 API calls and 1 hour of remote browser, no credit card; Starter then includes 50 calls and 10 browser hours a month

Lowest paid plan

Professional, $99 a month with 10,000 API calls and 500 remote browser hours

Key ambiguities

Enterprise pricing and what the dedicated cloud or on premise deployment costs.

Missing data

Enterprise rates.

Agentic Index verified 2026-09-27

Official pricing page

Related vendors

Alternatives to AgentQL

The closest documented capability profiles to AgentQL among browser and computer-use agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

Autotab8.0 / 14Adds documented Integrations & Tool Calling and Workflow Orchestration, among others
Stagehand7.0 / 14Adds documented Integrations & Tool Calling and Workflow Orchestration, among others
Bright Data Agent Browser4.5 / 14Adds documented Integrations & Tool Calling and Security, Identity & Governance
Notte9.5 / 14Adds documented Integrations & Tool Calling and Workflow Orchestration, among others
TinyFish7.5 / 14Adds documented Integrations & Tool Calling and Workflow Orchestration, among others
AGI, Inc.6.0 / 14Adds documented Workflow Orchestration and Human Oversight & Guardrails, among others

Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded

Compare AgentQL with other vendors →Compare all browser and computer-use agents →See where AgentQL ranks →

Data confidence: high

Head to head

AgentQL vs Lightpanda