Introducing /interact. Scrape any page, then let your agent take over to click, type, and extract data for you. Try it now →

[ INTERACT ]

[ PROMPT ]

[ SCRAPE ]

[ CODE ]

New: Prompt-based Interaction

Introducing
Interact

Scrape a page, then take actions on it. Describe what you want in natural language — or write code for full control.

luxestore.com/collections

LUXE

Search...

All

Bags

Shoes

Watches

Latest ▾

Leather Tote

$1,290

Running Shoes

$340

Silk Scarf

$180

Gold Watch

$4,200

‹

›

Agent

Agent Terminal

Interact

$firecrawl interact -p

>prompt: "search for leather bags"

>prompt: "filter by Casual style"

>prompt: "go to page 2"

>prompt: "extract all product cards"

>output: structured CSV (12 products)

[ 01 / 09 ]

How It Works

Scrape, then interact

Scrape any page, then interact with it using prompts or code.

Scrape the page

Get clean data from any URL. The response includes a scrapeId you'll use to interact.

firecrawl scrape https://example.com

Interact with prompts

Describe what you want in natural language. Click buttons, fill forms, extract data.

firecrawl interact -p "Click the login button"

Chain interactions

Each call reuses the session. State carries over so you can perform multiple actions in 1 session.

firecrawl interact -p "Fill email with test@example.com"
firecrawl interact -p "Submit the form"

✻Agent session

INTERACT

✻

Claude Code — Firecrawl toolkit ready

>Scrape all leather bag products with prices from luxstore.com

Works with prompts, Playwright code, and agent-browser CLI.

Read docs

[ 02 / 09 ]

Why Interact

Interact

Why interact?

The fastest way to go from scraping to taking action on any page.

Prompt-Based

Just describe what you want in natural language. Click buttons, fill forms, navigate pages — no selectors or Playwright code needed.

terminal

Full Code Control

Need precision? Write Playwright code in Node.js, Python, or Bash. The page object is ready — execute anything the browser can do.

your infrastructure

sandbox

Persistent Profiles

Save browser state across sessions with profiles. Stay logged in, preserve cookies, and pick up where you left off.

app.example.com

session 1

my-profile

auth

prefs

Session Chaining

Each interact call reuses the same session. Chain prompts to complete multi-step workflows — state carries over automatically.

firecrawl skill

Claude Code

Codex

OpenCode

[ 03 / 09 ]

CLI

The Interact Toolkit in your terminal

Scrape, interact, and stop — all from the terminal.

Terminal

$ firecrawl scrape https://example.com

{
  "title": "Example",
  "markdown": "# Hello\n...",
  "metadata": {
    "scrapeId": "<string>",
    "statusCode": 200,
    "sourceURL": "https://example.com"
  }
}

$ firecrawl interact -p "Click the sign up button"

{
  "success": true,
  "output": "I clicked the sign up button...",
  "liveViewUrl": "..."
}

$ firecrawl interact --stop

Session stopped.
Duration: 45.2s
Credits billed: 2

firecrawl

Scrape

firecrawl

Interact

firecrawl

Stop

[ 04 / 09 ]

Start Building

Integrations

Works with every agent framework

Connect Firecrawl Interact to your agent framework in minutes. Works out of the box with MCP.

Vercel AI SDK

Build AI apps with streaming and tool use

import { generateText } from "ai";
import { openai } from "@ai-sdk/openai";
import { createMCPClient } from "ai";

const client = await createMCPClient({
  transport: { type: "sse",
    url: "https://mcp.firecrawl.dev/sse" },
});

const { text } = await generateText({
  model: openai("gpt-4o"),
  tools: await client.tools(),
  prompt: "Scrape example.com and click the "
    + "login button on the page",
});

OpenAI Agents

Agentic workflows with tool calling

from agents import Agent, Runner
from agents.mcp import MCPServerSse

firecrawl = MCPServerSse(
    url="https://mcp.firecrawl.dev/sse",
    headers={
        "Authorization": "Bearer fc-YOUR-API-KEY"
    },
)

agent = Agent(
    name="Interact Agent",
    instructions="Scrape pages and interact "
        "with them for the user.",
    mcp_servers=[firecrawl],
)

result = await Runner.run(
    agent, "Scrape example.com, click the "
    "pricing tab and extract the plans"
)

Claude

MCP-native tool use with Anthropic

import anthropic

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    tools=[{
        "type": "mcp",
        "server_url":
          "https://mcp.firecrawl.dev/sse",
        "server_label": "firecrawl",
    }],
    messages=[{
        "role": "user",
        "content": "Scrape example.com and "
          "fill the contact form with "
          "test data, then submit it"
    }],
)

LangChain

Composable agents with LangGraph

from langchain_mcp_adapters.client \
    import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

client = MultiServerMCPClient({
    "firecrawl": {
        "url": "https://mcp.firecrawl.dev/sse",
        "transport": "sse",
    }
})

tools = await client.get_tools()
agent = create_react_agent(
    ChatOpenAI(model="gpt-4o"), tools
)
result = await agent.ainvoke({
    "messages": "Scrape news.ycombinator.com "
    "and extract the top 5 story titles"
})

Don't see your favorite framework? Email help@firecrawl.com

View docs

[ 05 / 09 ]

Use Cases

Go beyond static scraping

When scraping gives you the data, interact lets you take action on it.

E-commerce Monitoring

Scrape product pages and interact to add items to cart, check prices after login, or track inventory changes.

$ firecrawl interact -p "Add the first item to cart and check the total price"

Form Automation

Fill out multi-step forms, submit applications, or complete signup flows with natural language.

$ firecrawl interact -p "Fill the registration form with test data and submit"

Deep Data Extraction

Navigate paginated results, expand dropdowns, and click through tabs to extract data scraping alone can't reach.

$ firecrawl interact -p "Click next page and extract all product listings"

Authenticated Workflows

$ firecrawl interact -p "Log in and download the monthly report"

QA & Testing

Verify UI flows, check form validations, and test multi-step user journeys across pages.

$ firecrawl interact -p "Submit the form with invalid email and check the error message"

Multi-Step Workflows

Chain prompts to complete complex flows — search, filter, select, checkout — all in one session.

$ firecrawl interact -p "Search for laptops, filter by price under $1000, and extract the top 5 results"

[ 06 / 09 ]

Code Examples

Scrape, prompt, done

Scrape a page, then interact with prompts or code. Works with every SDK.

Prompt or code

Describe what you want in natural language, or write Playwright code for full control. Both use the same session.

Session chaining

Each interact call reuses the session. Chain prompts to complete multi-step flows — state carries over between calls.

Live view included

Every response includes a live view URL. Watch the browser in real time or embed the interactive view in your own UI.

[ SDK ]

import Firecrawl from '@mendable/firecrawl-js';

const firecrawl = new Firecrawl({ apiKey: "fc-YOUR-API-KEY" });

// Scrape the page
const scrape = await firecrawl.scrapeUrl("https://news.ycombinator.com");
const scrapeId = scrape.metadata.scrapeId;

// Interact with a prompt
const result = await firecrawl.interact(scrapeId, {
  prompt: "Extract the top 5 story titles",
});
console.log(result.output);

// Or run Playwright code
const result2 = await firecrawl.interact(scrapeId, {
  code: "return await page.title()",
  language: "node",
});
console.log(result2.result);

// Stop when done
await firecrawl.interactStop(scrapeId);

from firecrawl import FirecrawlApp

app = FirecrawlApp(api_key ="fc-YOUR-API-KEY")

# Scrape the page
scrape = app.scrape_url("https://news.ycombinator.com")
scrape_id = scrape["metadata"]["scrapeId"]

# Interact with a prompt
result = app.interact(scrape_id,
    prompt ="Extract the top 5 story titles"
)
print(result["output"])

# Or run Playwright code
result = app.interact(scrape_id,
    code ="return await page.title()",
    language ="node",
)
print(result["result"])

# Stop when done
app.interact_stop(scrape_id)

# Scrape the page
firecrawl scrape https://news.ycombinator.com

# Interact with prompts
firecrawl interact  -p "Extract the top 5 story titles"
firecrawl interact  -p "Click the first story link"
firecrawl interact  -p "Summarize this page"

# Or run code
firecrawl interact  -c "return await page.title()"

# Stop when done
firecrawl interact --stop

# 1. Scrape the page
curl  -X POST "https://api.firecrawl.dev/v2/scrape" \
   -H "Authorization: Bearer $FIRECRAWL_API_KEY" \
   -H "Content-Type: application/json" \
   -d '{"url": "https://news.ycombinator.com"}'

# 2. Interact with a prompt
curl  -X POST "https://api.firecrawl.dev/v2/scrape/SCRAPE_ID/interact" \
   -H "Authorization: Bearer $FIRECRAWL_API_KEY" \
   -H "Content-Type: application/json" \
   -d '{"prompt": "Extract the top 5 story titles"}'

# 3. Stop
curl  -X DELETE "https://api.firecrawl.dev/v2/scrape/SCRAPE_ID/interact" \
   -H "Authorization: Bearer $FIRECRAWL_API_KEY"

[ OUTPUT ]

// Output (prompt)
> Interacting with scraped page...

Here are the top 5 stories on Hacker News:
1. Mass Recall of Tesla Cybertrucks
2. Show HN: I built a real-time flight tracker
3. The Art of PostgreSQL
4. Why SQLite Is So Great for the Edge
5. Ask HN: What are you working on?

✓ Prompt executed successfully

Scrape a page → interact with it → stop when done. That's it.

[ 07 / 09 ]

FAQ

Frequently asked questions

Everything you need to know about Firecrawl Interact.

Interact

Pricing

Usage

FOOTER

The easiest way to extract
data from the web

Backed by

Y Combinator

Linkedin Github YouTube

SOC II · Type 2

AICPA

SOC 2

X (Twitter)

Discord

IntroducingInteract

Scrape, then interactScrape, then interact

Why interact?Why interact?

The Interact Toolkit in your terminalThe Interact Toolkit in your terminal

Works with every agent frameworkWorks with every agent framework

Go beyond static scrapingGo beyond static scraping

Scrape, prompt, doneScrape, prompt, done

Frequently asked questionsFrequently asked questions

Introducing
Interact

Scrape, then interact

Why interact?

The Interact Toolkit in your terminal

Works with every agent framework

Go beyond static scraping

Scrape, prompt, done

Frequently asked questions