Home/AI coding agent cost
AI Coding Agent Cost Calculator
What running a coding agent costs on pay-as-you-go API pricing, and when a flat plan is cheaper. A typical workload costs about $180 a month on Claude Sonnet 5.5, and from $6.91 (GPT-6 Luna) to $691 (GPT-6 Astra) across 12 models. Prices verified October 2026.
Estimate Your Monthly Agent Bill
Same workload on other models (per month)
Flat plans that include Claude Sonnet 5.5, against $180/month of API usage
| GitHub Copilot · Pro | $10 | ≈ $175 total with $165 of extra credits |
| Claude Code · Claude Pro | $20 | Cheaper than the API, if its usage limits cover you |
| Cursor · Pro | $20 | Cheaper than the API, if its usage limits cover you |
| Windsurf · Pro | $20 | Cheaper than the API, if its usage limits cover you |
| GitHub Copilot · Pro+ | $39 | ≈ $149 total with $110 of extra credits |
| Cursor · Pro+ | $60 | Cheaper than the API, if its usage limits cover you |
| GitHub Copilot · Max | $100 | Included credits cover your month |
| Claude Code · Claude Max 5x | $100 | Cheaper than the API, if its usage limits cover you |
| Claude Code · Claude Max 20x | $200 | The API is cheaper at your volume |
| Cursor · Ultra | $200 | The API is cheaper at your volume |
An estimate, not a quote: real sessions vary a lot with task size, how many files the agent reads and how long a conversation runs before it is cleared.
What the Vendors Report
$150–250 per developer per month; 90% of users stay under $30 per active day. Enterprise deployments on API billing.
“Across enterprise deployments, the average cost is around $13 per developer per active day and $150-250 per developer per month, with costs remaining below $30 per active day for 90% of users.”Source ↗
Per developer, on Business and Enterprise credit billing. Varies with model, parallel instances, automations and fast mode.
“On average, Codex costs approximately $100–$200 per developer per month”Source ↗
Total usage for daily Agent users; power users running multiple agents or automation often pass $200.
“Daily Agent users: Typically $60–$100/mo total usage”Source ↗
Agent usage is billed per token at each model's price. Plans include $15 (Pro), $70 (Pro+) or $200 (Max) of credits a month.
“Copilot usage is measured in AI credits—a usage-based billing unit where 1 AI credit = $0.01 USD.”Source ↗
Flat Plans for Coding Agents
Individual plans, monthly billing. Most vendors cap usage per time window instead of publishing a dollar amount; GitHub Copilot is the exception, with credits billed at each model's price.
| Plan | Per month | What you get |
|---|---|---|
| GitHub Copilot · Pro | $10 | 1,500 AI credits ($15) a month; no Opus |
| Gemini CLI · Google AI Pro | $19.99 | 1,500 model requests a day |
| Claude Code · Claude Pro | $20 | 5-hour and weekly limits shared with chat; extra usage at API rates |
| Codex · ChatGPT Plus | $20 | 15–160 GPT-6.1 Sol local messages per 5 hours; weekly limits may apply |
| Cursor · Pro | $20 | Included usage pools; third-party models drawn down at API prices |
| Kiro · Pro | $20 | 1,000 credits a month, extra credits $0.04 each; lists GPT-5.6 and Claude 5 models |
| Windsurf · Pro | $20 | Daily and weekly quota; extra usage at API pricing |
| Gemini CLI · Code Assist Standard | $22.8 | 1,500 model requests a day ($19 billed annually) |
| GitHub Copilot · Pro+ | $39 | 7,000 AI credits ($70) a month, including Opus |
| Cursor · Pro+ | $60 | Larger included pools; on-demand usage at API rates |
| Claude Code · Claude Max 5x | $100 | 5x Pro usage per 5-hour session, weekly limit |
| Codex · ChatGPT Pro | $100 | No 5-hour limit for now; $200 and $500 tiers raise the allowance |
| GitHub Copilot · Max | $100 | 20,000 AI credits ($200) a month, priority access |
| Claude Code · Claude Max 20x | $200 | 20x Pro usage per 5-hour session, weekly limit |
| Cursor · Ultra | $200 | Largest included pools; on-demand usage at API rates |
Bring-your-own-key tools such as Cline and Roo Code charge nothing on top: you pay the API price shown in the calculator. Product names are trademarks of their owners.
Cost by Tool
Each tool's flat plans next to the API cost of the same typical workload (25 sessions a day, 15 active days) on the models those plans list.
Claude Code cost
Plans: Claude Pro $20/month, Claude Max 5x $100/month, Claude Max 20x $200/month. The vendor reports ~$13 per active day: $150–250 per developer per month; 90% of users stay under $30 per active day. Enterprise deployments on API billing.
- Claude Haiku 4.5 via API$69.08/month
- Claude Sonnet 5.5 via API$180/month
- Claude Opus 5.5 via API$268/month
- Claude Fable 5.1 via API$554/month
Codex cost
Plans: ChatGPT Plus $20/month, ChatGPT Pro $100/month. The vendor reports ~$100–200 per month: Per developer, on Business and Enterprise credit billing. Varies with model, parallel instances, automations and fast mode.
- GPT-6 Luna via API$6.91/month
- GPT-6.1 Sol via API$103/month
- GPT-6 Astra via API$691/month
Cursor cost
Plans: Pro $20/month, Pro+ $60/month, Ultra $200/month. The vendor reports $60–100 per month: Total usage for daily Agent users; power users running multiple agents or automation often pass $200.
- Gemini 3.8 Flash via API$48.29/month
- Gemini 3.1 Pro via API$133/month
- Claude Sonnet 5.5 via API$180/month
- Grok 4.7 via API$227/month
- Claude Opus 5.5 via API$268/month
GitHub Copilot cost
Plans: Pro $10/month, Pro+ $39/month, Max $100/month. The vendor reports 1 AI credit = $0.01: Agent usage is billed per token at each model's price. Plans include $15 (Pro), $70 (Pro+) or $200 (Max) of credits a month.
- Gemini 3.8 Flash via API$48.29/month
- Claude Haiku 4.5 via API$69.08/month
- Claude Sonnet 5.5 via API$180/month
- Claude Opus 5.5 via API$268/month
- Claude Fable 5.1 via API$554/month
Gemini CLI cost
Plans: Google AI Pro $19.99/month, Code Assist Standard $22.8/month.
- Gemini 3.8 Flash via API$48.29/month
- Gemini 3.1 Pro via API$133/month
How the Estimate Works
- One session uses the token mix of Anthropic's example session: about 997K tokens, 94% of them read from the prompt cache.
- Claude 4.7 and later models count about 30% more tokens for the same text, so their sessions are scaled up.
- Cache writes cost 1.25x the input price on Anthropic and OpenAI models; other providers list no cache-write charge.
- Checked against vendor averages: the typical preset on Claude Sonnet 5.5 costs $11.97 per active day, close to the $13 Anthropic reports.
AI Coding Agent Cost — FAQ
On API pricing, a typical workload (25 agent sessions a day, 15 active days a month) costs about $180 a month on Claude Sonnet 5.5, or $11.97 per active day. Heavy use (60 sessions a day) reaches about $431. Vendors report similar figures: Anthropic says Claude Code averages around $13 per developer per active day ($150–250 a month), OpenAI says Codex averages $100–200 per developer per month, and Cursor says daily Agent users typically use $60–100 a month.
Usually, if its limits cover you. Flat plans cost $10.00 to $200 a month, while the same typical workload costs $180 on the Claude Sonnet 5.5 API. The catch is usage limits: Claude and ChatGPT plans cap use per 5-hour window and per week, and GitHub Copilot includes a fixed dollar amount of credits (for example $70 on Pro+) billed at model prices, with extra credits at $0.01 each.
An agent re-sends the whole conversation, including files it has read and tool results, on every step. In Anthropic's example session about 997K tokens were processed and 94% of them were read from the prompt cache, which costs a tenth of the normal input price or less. Without caching the same session would cost several times more.
At the same workload, GPT-6 Luna is the cheapest model compared here at about $6.91 a month, and GPT-6 Astra the most expensive at about $691. Cheaper models may need more attempts on hard tasks, so the cheapest per session is not always the cheapest per finished task.
Each session uses the token mix from Anthropic's documented example: 1,200 input, 5,300 output, 940,000 cache-read and 50,000 cache-write tokens. Claude 4.7 and later models count about 30% more tokens for the same text, so their sessions are scaled up. Cache writes cost 1.25x the input price on Anthropic and OpenAI models. Each provider's official API price is applied, verified October 2026.