Agentql
AgentQL is a suite of tools designed to connect AI agents to the web, enabling web interaction and structured data extraction through a specialized query language.
Verdict
Common use cases
- Extract competitor pricing from public sites
- Pull job listings into a hiring tracker
- Monitor product availability across vendors
- Scrape event schedules for calendar imports
- Automate data entry from partner portals
Integration
- Vendor
- Agentql
- Category
- other
- Auth
- API_KEY
- Tools
- 3
- Composio slug
agentql
Tools
- Create Remote Browser Session
Tool to create a remote browser session. use when you need to run browser automation on remote infrastructure.
- Get Usage Details
Tool to retrieve subscription usage details for the current api key, including usage limits and billing period. use after confirming api connection is valid.
- Query Data
Tool to query structured data as json from a web page using an agentql query or natural language prompt. use after defining your query or prompt and a url or html.
Setup
Setup guide
- 11. Open your Switchy workspace and navigate to Settings > Integrations > MCP Servers. 2. Click 'Add MCP Server' and select AgentQL from the list. 3. Go to the AgentQL dashboard at agentql.com, sign in, and generate a new API key under Account > API Keys. 4. Paste the API key into Switchy's connection form and click 'Connect'. 5. Switchy will test the connection by calling the Get Usage Details tool to confirm your key is valid. 6. Once connected, open any Space and type '@AgentQL query data from https://example.com/pricing' to test a live extraction. 7. The MCP will return structured JSON you can review in the chat thread. 8. For complex scraping tasks, use the Create Remote Browser Session tool to run automation on AgentQL's infrastructure instead of your local machine.
What teammates see: by default, memories from Agentql are scoped to the Space (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Spaces have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
Extract Pricing Table
@AgentQL query data from https://example.com/pricing and return a JSON array with fields for plan_name, monthly_price, and feature_list.Open in a Space →
Monitor Job Postings
@AgentQL query data from https://company.com/careers and extract job_title, location, and apply_url for all open positions.Open in a Space →
Check Stock Availability
@AgentQL query data from https://store.com/product/12345 and return in_stock status and current_price.Open in a Space →
Scrape Event Calendar
@AgentQL query data from https://venue.com/events and return event_name, date, and location for all upcoming events.Open in a Space →
Check API Usage Limits
@AgentQL get usage details for my API key and show me remaining queries and reset date.Open in a Space →
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@agentql extract the top 5 product names and prices from https://example-store.com/laptops as JSON
AgentQL navigated to the URL and extracted structured data using a natural-language query. The response includes 5 product objects, each with a 'name' and 'price' field, returned as clean JSON. For example: {'name': 'ThinkPad X1 Carbon', 'price': '$1,299'}. This illustrative output shows how the MCP converts unstructured HTML into queryable data without writing CSS selectors or XPath.
This example highlights AgentQL's core strength: turning any public webpage into structured JSON via natural language. Requires a valid API key. The MCP does not store or cache scraped data—each query is a fresh fetch. Be mindful of rate limits on your subscription tier and respect the target site's robots.txt.
@agentql create a remote browser session and navigate to https://dashboard.example.com, then screenshot the analytics panel
AgentQL provisioned a headless Chrome instance in the cloud, loaded the specified URL, and captured a screenshot of the viewport. The illustrative response confirms session ID, browser version, and a base64-encoded PNG of the dashboard. This example demonstrates remote automation—useful when the target site requires JavaScript rendering or interactive elements that static scraping cannot handle.
Remote browser sessions consume more API credits than simple queries and introduce latency (typically 3-8 seconds for session spin-up). Useful for SPAs, login-gated pages, or dynamic content. Sessions auto-terminate after inactivity. Ensure your API key has browser-automation scope enabled.
@agentql check my current usage limits and tell me how many queries I have left this billing cycle
AgentQL retrieved subscription metadata for the authenticated API key. The illustrative response shows: 'You have used 1,240 of 5,000 queries this month. Your billing period resets on March 1. Browser sessions: 18 of 50 used.' This example pairs the MCP's usage tool with the AI's ability to interpret quotas and suggest pacing strategies if you're approaching limits.
This read-only tool helps teams monitor API consumption before hitting hard limits. The MCP does not modify your subscription or trigger billing actions. If usage data is stale (cached up to 5 minutes), the AI will note that in its response. Useful for cost-aware workflows.
Use-case deep-dives
When scraping beats manual price checks for a 3-person growth team
A growth team at a B2B SaaS startup needs to track competitor pricing changes weekly across 12 rivals. Agentql's Query Data tool pulls structured JSON from pricing pages without writing selectors—you describe what you want in natural language and it extracts the data. The remote browser session handles dynamic pages that load prices via JavaScript. This works when your target sites don't change layout every week and you're checking under 50 pages per run. If a competitor redesigns their site, you rewrite the prompt instead of debugging XPath. The API key auth means any teammate can run the query from Switchy without sharing credentials. Best fit: recurring research tasks where the data structure is predictable but the page markup isn't.
Extracting help-center content when you can't get API access
A 6-person support team wants to pull FAQ content from a vendor's help center into their internal knowledge base, but the vendor has no API. Agentql's natural-language query lets you describe the article structure—title, body, category—and get clean JSON back without parsing HTML yourself. The remote browser session means you're not blocked by rate limits or IP bans that hit local scrapers. This is the right call when you're doing a one-time migration or monthly sync of under 500 articles. If you need real-time updates or the help center requires login flows with CAPTCHA, Agentql won't solve it. The 3-tool limit means you're running simple extract-and-save workflows, not complex multi-step automations. Best fit: one-off or low-frequency content pulls from public pages.
When scraping LinkedIn alternatives beats paying for enrichment APIs
A 4-person sales team wants to enrich inbound leads with job titles and company info from public profiles on sites like Crunchbase or AngelList. Agentql's Query Data tool extracts structured fields—name, role, funding stage—without maintaining a scraper codebase. The remote browser session handles sites that block headless Chrome. This works when you're enriching 20-50 leads per week and the profiles are public. If you're doing 500+ lookups daily or need historical data, a paid enrichment API is faster and more reliable. The natural-language query means non-technical teammates can adjust what fields to pull without writing code. Best fit: low-volume lead research where you're trading engineer time for per-lookup cost.
Frequently asked
What does the Agentql MCP do in Switchy?
It lets your team extract structured data from any webpage or run browser automation tasks without writing scraper code. You give it a URL and describe what data you want in plain English, and it returns clean JSON. Useful for pulling competitor pricing, monitoring job boards, or grabbing product specs from sites without APIs.
Do I need an Agentql API key to use this MCP?
Yes. You authenticate with an Agentql API key, which you generate in your Agentql dashboard. Each key is tied to a usage quota and billing period. The MCP includes a tool to check your remaining quota, so you know when you're approaching limits before a scrape job fails mid-run.
Can the Agentql MCP interact with pages that require login?
Yes, via the remote browser session tool. You can spin up a headless browser, log in manually or programmatically, then run queries on authenticated pages. This works for scraping internal dashboards, competitor tools behind paywalls, or any site that blocks unauthenticated requests. Sessions persist until you close them.
How is this different from just using Agentql's API directly?
The MCP wraps Agentql's API so your team can run scrapes from inside Switchy conversations without switching to Postman or writing Python scripts. You describe the task in chat, the MCP handles the API calls, and results appear inline. Faster for ad-hoc research; less flexible than a custom integration for production pipelines.
Who on my team should connect the Agentql MCP?
Whoever owns your Agentql account and has the API key. Usage counts against that key's quota, so connect it under a shared service account if multiple people will run scrapes. If you're on a tight quota, consider restricting access to avoid surprise overages mid-month.