Costs

Startica itself is free and open source. The main running cost is the AI provider you pick, Claude (Anthropic) or OpenAI, billed to your own account. The only other paid service is the optional X API, billed by X if you add your own token. Reddit, Hacker News, Bluesky, Mastodon and Discourse are read for free, and Google Analytics, Google Search Console, Bing Webmaster Tools, Supabase and Stripe access don’t cost anything extra through Startica; they use your existing accounts’ normal quotas. Search Console and Bing Webmaster Tools, and their APIs, are free. The agent readiness checks and off-site presence use no AI and cost nothing.

This page lists the prices Startica uses to estimate cost, gives rough figures for common actions, and explains the monthly budget cap.

Prices Startica uses

Startica estimates the cost of every call from the token counts the API returns and the price table in packages/core/src/ai/models.ts. Prices are in USD per million tokens, each provider’s first-party API rates as checked when the code was written.

Model	Input	Output	Cache write	Cache read (cached input)
Claude Fable 5.1 (claude-fable-5-1)	$10.00	$50.00	$12.50	$0.25
Claude Opus 5.5 (claude-opus-5-5)	$4.00	$20.00	$5.00	$0.20
Claude Sonnet 5.5 (claude-sonnet-5-5)	$2.00	$10.00	$2.50	$0.20
Claude Haiku 5.5 (claude-haiku-5-5)	$0.10	$0.50	$0.125	$0.01
GPT-6 Astra (gpt-6-astra)	$10.00	$50.00	$12.50	$1.00
GPT-6.1 Sol (gpt-6.1-sol)	$2.00	$10.00	$2.50	$0.10
GPT-6 Luna (gpt-6-luna)	$0.10	$0.50	$0.125	$0.01
GPT Image 2.5 Flare (gpt-image-2.5-flare)	$5.00	$30.00	$6.25	$1.25
GPT Image 2.5 Sunburst (gpt-image-2.5-sunburst)	$5.00	$30.00	$6.25	$1.25

• Cache write is 1.25 times the input price at both providers. Cache read is what repeated prompt prefixes cost; see prompt caching.
• Web search, used by research calls and AI visibility answers, adds $10 per 1,000 searches ($0.01 each) on top of tokens, at both providers. Claude’s web fetch costs only the tokens of the fetched page.
• Thinking (reasoning) tokens count as output tokens. Higher Effort settings usually mean more of them.
• Image models are billed per token too: the input is the prompt’s text, the output is the picture. See Ad pictures.

The providers’ actual prices are on the Claude API pricing page and the OpenAI API pricing page. If they change and the code hasn’t been updated yet, Startica’s estimates drift from your real bill. Your provider’s invoice is always the authority.

Rough cost of common actions

These are estimates worked out from the default models, typical prompt sizes in the code and assumed output lengths. They aren’t measurements. Your numbers depend on your product brief, thread lengths, effort settings and how much gets cached. Check the AI calls table on the System page for real figures.

Action	Task and default model	Rough cost
Test job (Run test job)	ping, Haiku or Luna	Well under $0.001
Score one conversation	score, Haiku or Luna	About $0.0005
Score one source’s full share of a scan (60 new conversations, the per-source cap)	score, Haiku or Luna	About $0.03
Draft one reply	draft, Sonnet or Sol	$0.01 to $0.03
Write one Content Studio piece	draft, Sonnet or Sol	$0.01 to $0.05
Build a voice profile	strategy, Opus or Sol	$0.10 to $0.25
Build an Intel brief (research plus synthesis)	research + strategy, Opus or Sol	$0.40 to $1.00
Plan or replan the roadmap	strategy, Opus or Sol	$0.10 to $0.30
Plan an ad campaign	strategy, Opus or Sol	$0.10 to $0.30
Rewrite one ad	draft, Sonnet or Sol	$0.01 to $0.03
Suggest three ad creative ideas	draft, Sonnet or Sol	$0.01 to $0.03
Draw one ad picture	image, GPT Image 2.5 Flare at medium quality	About $0.05
Landing, onboarding or SEO audit	strategy, Opus or Sol	$0.10 to $0.25
GEO audit (includes web research)	research + strategy, Opus or Sol	$0.30 to $0.70
Suggest AI visibility questions	draft, Sonnet or Sol	$0.01 to $0.05
Draft one fix on Opportunities (titles and descriptions, JSON-LD, llms.txt or an FAQ answer)	draft, Sonnet or Sol	$0.01 to $0.03
Write it in Content Studio on Opportunities (one blog article)	draft, Sonnet or Sol	$0.01 to $0.05
Ask one AI visibility question (answer with up to 3 searches, then read it)	answer + score, Sonnet + Haiku or Sol + Luna	About $0.06
An AI visibility run of 20 questions	answer + score	About $1.30, capped at $2 by default
Accessibility: suggest a fix for one issue, Ask AI on one Needs review item, or summarize a site	draft or vision, Sonnet or Sol	About $0.01 each
Accessibility: review the alt text of 50 images	vision, Sonnet or Sol	About $0.25 (about $0.005 an image)
Security: explain and fix one finding, or triage one reported problem	draft, Sonnet or Sol	About $0.02 each
Security: import another tool’s report	draft, Sonnet or Sol	About $0.10 for a long report

These figures are worked out for Claude’s defaults. OpenAI’s defaults use the same prices for scoring and drafting and a cheaper model (GPT-6.1 Sol) for strategy and research, so those actions usually cost less on OpenAI.

A rough month

As an illustration only: one product, around 300 conversations scored a day, 5 drafts a day, a weekly replan and a couple of audits comes to roughly $10 to $20 a month with the default models. The default budget of $50 leaves room for that plus rebuilding briefs and writing content.

What moves the total most:

• Scoring volume. Startica scans every hour and scores up to 60 new conversations per product from each source per scan (X at most 50). Broad keywords or many watched subreddits mean more scoring. Each one is cheap, but they add up.
• Auto-drafting. Every conversation that clears your relevance and intent thresholds gets a draft automatically. Raising the thresholds means fewer, better drafts. See Finding conversations.
• AI visibility runs. About $0.06 a question, so weekly runs of 20 questions add about $5 to $6 a month. They’re off until you turn them on, and each run has its own cap. See AI visibility.
• Model choices. Switching score to Sonnet or GPT-6.1 Sol multiplies scoring cost by about 20. Switching strategy from Opus to Fable or GPT-6 Astra multiplies it by 2.5.

AI visibility

AI visibility asks an AI model your customers’ questions with web search on, so it costs more per call than most features: the search results arrive as input tokens, and each search costs $0.01.

Before each run, the Run now dialog shows an estimate: the number of active questions times the estimated cost of one question. One question is assumed to be an answer of 12,000 input and 900 output tokens on the answer model, 3 web searches, and a reading of 2,500 input and 250 output tokens on the score model. On the default models (Sonnet and Haiku, or GPT-6.1 Sol and GPT-6 Luna), that’s about $0.063 a question:

Questions	Estimate
12	About $0.76
20	About $1.27
100	About $6.34, more than the default cap

Every run has a cost cap per product: $2 by default, from $0.10 to $50, set in Settings on the AI visibility screen. Before each question, Startica adds what the run has spent so far to the estimate for one more question, and stops the run if that would pass the cap, keeping the answers it has. A question that costs more than its estimate can take the run a few cents over. The monthly budget applies on top: when it’s reached, the run stops and fails.

Weekly runs are off by default. With them on, each product with active questions runs every Monday, each run under its cap. Suggesting questions is a single draft call of a few cents. See What a run costs.

After each run, Off-site presence reads some of the pages the answers cited. That read uses no AI and costs nothing.

Ad pictures

Ad creatives can have their picture drawn by OpenAI’s image model, on the image task. It needs an OpenAI API key even when Claude is your provider; see The image task. Uploading your own picture, or using no picture, costs nothing, and so does rendering the sizes: that happens on your computer.

Each picture is drawn once, at 2048×2048, and every size the platform takes is cropped from it, so one picture costs the same whichever platform it’s for. Before you draw, the creative shows an estimate, for example “about $0.05 at medium quality”. It assumes 400 input tokens of prompt and, for the output, OpenAI’s per-image token counts scaled to 2048×2048 (4 megapixels):

Quality	Output tokens	Estimate
low	784	About $0.03
medium (default)	1,756	About $0.05
high	7,024	About $0.21
xhigh	12,488	About $0.38
max	28,096	About $0.84

For GPT Image 2.5 Sunburst the estimate is 1.5 times Flare’s, because Sunburst uses more output tokens for the same picture. After drawing, the real cost from the API’s token counts is added to the creative (shown on its card) and to the campaign’s total on the Creatives tab, and logged under AI calls. Every picture counts against the monthly budget. Draw again costs another picture, and Startica asks you to confirm first. Suggest ideas doesn’t draw anything: the ideas wait until you click Draw it.

X API (optional)

If you add an X API bearer token, X bills your X developer account for every post the API returns. Startica estimates about $0.005 per post read.

Setting	Default	Cost at the default
Most post reads per day	100	About $0.50 a day, at most about $15 a month
Scan X at most every … hours	6	Up to 4 searches a day, each reading at most 50 posts, all within the daily cap

Searches, pasted X links and engagement checks on your X posts all count toward the daily cap, which resets at midnight UTC. The X cost isn’t part of the monthly AI budget below and isn’t shown on the System page; the X API (optional, paid) card in Settings → Channels shows today’s reads. X posts that Startica finds are scored and drafted by the AI like any other conversation, which costs AI budget as usual.

Without a token, X costs nothing: you paste X posts yourself.

The monthly budget

Startica enforces its own monthly cap on AI spend, separate from any limit you set with your provider. It covers both providers together.

Setting	Detail
Where	Settings → General → AI → Add API key or Change keys or budget, field Monthly budget (USD); also the welcome wizard’s Connect AI step
Default	$50
Allowed range	$1 to $10,000
Period	The calendar month, in your computer’s local time

What happens at the cap

Before every AI call, to either provider, Startica adds up the estimated cost of all calls this month from its own log (ai_runs). If the total has reached the budget:

1. The call isn’t made. No tokens are spent.
2. A blocked entry is logged under AI calls, with “Monthly AI budget reached ($X of $Y)”.
3. The job fails with the same message. You see it under Job runs on the System page, and on the page that started the job.

Startica doesn’t switch to a cheaper model when you’re near the cap. It runs normally until the cap, then blocks every AI call until the 1st of next month or until you raise the budget. Non-AI work keeps running: scanning feeds, syncing metrics, checking engagement on your posts. The monthly budget doesn’t limit X API reads; those have their own daily cap. New conversations still arrive, but scoring them fails while the cap is in effect.

The check happens before each call, so a call that’s already running when the cap is reached finishes normally. Spend can end up slightly over the budget.

Where to see spend

Where	What it shows
System → AI spend this month	This month’s estimated total and your budget, for example $3.2140 of $50. It counts only Startica’s calls. Links under it open the usage page of each provider you’ve saved a key for (platform.claude.com, platform.openai.com), which shows your account’s actual usage and bill, including other apps.
System → AI calls	The latest 20 calls: task, model, status, tokens in and out, estimated cost, latency
Settings → General → AI	The provider in use, whether each provider’s key is saved, and your budget
The Claude Console or OpenAI platform usage and billing pages	What your provider actually charged, across everything that uses your key

To dig further, the full history is in the ai_runs table of the local database. See Data folder.

Runs of yarn eval (for contributors) use a throwaway database. Their cost is printed in the terminal and doesn’t count toward the app’s monthly total, though your provider still bills it.