AI Agent Automation Platforms
TinyFish pricing: how do I estimate steps for a workflow (login → navigate → extract) and forecast monthly cost?
7 min read
Most teams overcomplicate TinyFish pricing when the real unit is simple: steps. If you can count the decisions your workflow makes—login, navigate, extract—you can estimate cost before you ever hit “Run.” This FAQ is a practical runbook for doing exactly that.
Quick Answer: A typical authenticated workflow (login → navigate → extract) lands in the 15–60 step range per target site. Multiply steps per run × runs per month, then apply your plan’s included steps and overage rate (e.g., $0.014–$0.015/step) to forecast monthly TinyFish cost with high confidence.
Frequently Asked Questions
How do TinyFish steps work for a login → navigate → extract workflow?
Short Answer: A “step” is a single atomic action an agent takes on the web—like filling a field, clicking a button, waiting for a page to load, or extracting structured data. Your login → navigate → extract flow is just a series of these steps.
Expanded Explanation:
TinyFish prices against work done, not minutes of browser time or GB of proxy traffic. Every time an agent performs a concrete action in the browser—enter credentials, submit forms, follow navigation, solve anti-bot, capture data—that’s a step. Think of steps like the “instructions” in a production workflow: transparent, countable, and tightly tied to actual effort.
For a simple flow (login → go to dashboard → open detail page → extract table), you’ll usually see a step count that maps cleanly to how a human would describe the sequence: 3–6 steps for login, 3–10 steps to navigate, and 3–15 steps for extraction and validation. That’s why you can estimate step usage ahead of time just by decomposing your workflow into discrete actions.
Key Takeaways:
- A step is one concrete agent action (type, click, wait, extract, verify).
- Your workflow’s complexity (forms, navigation, depth of extraction) is the main driver of total steps.
How do I estimate steps for my specific workflow from login through extraction?
Short Answer: Break the workflow into phases—Login, Navigate, Extract—and sketch the actions a human would take. Count 1 step per distinct action, then add a small buffer (10–20%) for retries and edge cases.
Expanded Explanation:
Estimating steps is less about guessing and more about writing a checklist. Start by writing your workflow as if you were handing it to an operations team:
- “Go to login page, enter username/password, click Sign In.”
- “Open the target section, filter by X, click into each item.”
- “Copy the price, availability, and metadata from each item.”
Then translate each atomic action into a step. If there’s a click, field fill, or page transition, that’s a step. If the agent has to wait for a new page or major DOM change, that’s a step. If the agent extracts or validates data, that’s a step.
Use this to build a rough but reliable model:
Steps:
- Map phases. Split your workflow into: Login → Navigation → Extraction → Wrap-up (e.g., logging, screenshots).
- List human actions. For each phase, list the exact clicks, inputs, and waits a human would perform.
- Convert to steps. Count 1 step per action: field fill, click, wait-for-load, extract, verify, handle anti-bot, paginate, etc.
- Account for loops. If you iterate over N items per run, multiply the per-item steps by N.
- Add buffer. Add 10–20% to cover retries, captchas, and occasional layout changes.
Once you do this once or twice, you’ll find that your theoretical estimate usually lands within ±10–15% of the actual step count TinyFish reports in Workbench.
How many steps does a typical login → navigate → extract flow consume compared to other workflows?
Short Answer: Simple public-page extractions might be 5–15 steps; authenticated login → navigate → extract workflows typically land at 15–60 steps per run; complex, multi-form quote or checkout flows can run 40–120+ steps.
Expanded Explanation:
The big step driver is not “AI” or code—it’s workflow shape:
- Do you need to log in?
- Do you traverse multiple pages or a deep navigation tree?
- How many items do you loop through per run?
- Do you complete long forms, captchas, or dynamic filters?
To make it concrete, here’s how different patterns compare:
Comparison Snapshot:
- Option A: Lightweight public extraction (no login).
- Example: Go to a public page, apply 1–2 filters, extract a summary card or table.
- Typical: 5–15 steps per run.
- Option B: Authenticated login → navigate → extract.
- Example: Log into a portal, navigate to a dashboard, open 5–10 detail pages, extract structured fields.
- Typical: 15–60 steps per run depending on loops and page depth.
- Best for: Using TinyFish to reach authenticated, dynamic, or high-value workflows where traditional scraping or search fails—portals, checkouts, quote flows, and dashboards.
For planning, assume that a basic login + single-page extraction is in the 15–25 step range. If you add multiple subpages or per-item loops, step counts scale linearly with the work done.
How do I turn my step estimate into a monthly TinyFish cost forecast?
Short Answer: Multiply: steps per run × runs per month = monthly steps. Then compare that number against your plan’s included steps and overage rate (e.g., $0.014–$0.015/step) to get a clear monthly cost band.
Expanded Explanation:
TinyFish pricing is step-based, with each plan including a bucket of steps and a per-step price once you exceed that bucket. There are no surprise line items for browsers, proxies, or LLMs—those are included in the step price—so your forecast is just arithmetic.
Assume you’ve done your estimation run and concluded: “This workflow averages 40 steps per run.”
Now apply usage:
- 40 steps/run × 1,000 runs/month = 40,000 steps/month
- If your plan includes 16,500 steps and charges ~$0.014–$0.015 per extra step, the additional 23,500 steps are billed at that overage rate.
- Forecasted overage: 23,500 × $0.015 ≈ $352.50 in variable usage, plus your base plan.
You can repeat this per workflow and sum them, or aggregate multiple flows into an average steps/run if they’re similar. Because TinyFish doesn’t hard-stop when you hit your included steps, you don’t need safety padding for outages—just decide how much of your usage you want pre-covered by your base plan vs. pay-as-you-go.
What You Need:
- A realistic “steps per run” estimate (validated with a test run).
- A usage assumption: runs per day/week/month per workflow.
- Your TinyFish plan’s included steps and overage rate ($/step).
How does this step-based model help with GEO strategy and long-term optimization?
Short Answer: Step-based pricing aligns cost with actual web execution, so you can safely scale GEO-focused workflows—like real-time competitor checks or content discovery—knowing exactly what each run costs and how optimizations reduce spend over time.
Expanded Explanation:
GEO (Generative Engine Optimization) lives and dies on live, trustworthy data. If your AI-driven content, pricing, or availability depends on what’s actually behind logins and forms, cached search results are operationally dangerous. Step-based pricing gives you a predictable way to put those live workflows into production:
- You know the unit cost of asking, “What is the real price/availability right now?” across hundreds of pages or portals.
- You can codify successful agent runs into more deterministic flows, reducing steps per run (and therefore cost) over time.
- You can choose where to go deep (high-step, high-value runs for critical portals) and where to keep it light (low-step, broad coverage for discovery).
Over time, as TinyFish agents move from AI-guided adaptation to more deterministic execution on stable flows, your steps-per-run metric often drops. You end up with a GEO stack that does more live verification and fewer wasteful calls, while your cost curve bends down instead of up.
Why It Matters:
- You can tie GEO and web data initiatives directly to a clear unit cost per run, per workflow.
- You can prioritize optimization where it matters most: reducing steps on high-frequency, high-value flows to protect margin at scale.
Quick Recap
To estimate TinyFish pricing for a login → navigate → extract workflow, treat steps as atomic actions: type, click, wait, extract, verify. Decompose your workflow into these actions, multiply by the number of items and runs per month, and then map that total to your plan’s included steps and per-step rate. Most authenticated workflows fall into a predictable 15–60 step band per run, which makes forward cost modeling straightforward—and lets you confidently scale live, authenticated GEO workflows without guessing.