AI coding tool pricing in 2026 is a mess. Credits, quotas, premium requests, per-token API billing, and daily caps all coexist — often within the same product. Two tools that both cost “$20/month” can produce wildly different monthly bills depending on how you use them. Cursor changed its billing model in mid-2025 and frustrated thousands of developers. Windsurf overhauled its pricing again on March 19, 2026. GitHub Copilot introduced premium request metering that makes heavy agent use surprisingly expensive.
This page cuts through the noise. We break down what every AI coding tool actually costs, what you get at each tier, where the hidden charges live, and which tool delivers the best value for your situation. We update this guide monthly as prices change.
Master Pricing Table
| Tool | Free Tier | Individual Plan | Team Plan | Enterprise Plan | Billing Model |
|---|---|---|---|---|---|
| Cursor | 2,000 completions + 50 slow premium requests | Pro: $20/month ($16 annual) | Teams: $40/user/month ($32 annual) | Custom (pooled usage, SSO, SCIM, audit logs) | Credit pool + pay-as-you-go overages |
| GitHub Copilot | 2,000 completions + 50 premium requests | Pro: $10/month ($100/year) / Pro+: $39/month | Business: $19/user/month | Enterprise: $39/user/month (requires $21/user Enterprise Cloud) | Premium request allowance + $0.04/request overage |
| Claude Code | Included in free Claude tier (limited) | Pro: $20/month (via Claude Pro) | Team: $30/user/month | Custom pricing | Per-token API or subscription-based allowance |
| Windsurf | Generous daily quota; unlimited Tab completions | Pro: $20/month | Teams: $40/user/month | Custom (200+ seats, SSO, RBAC) | Daily/weekly quota system (since March 19, 2026) |
| Gemini Code Assist | Free for individuals | Standard: $19/user/month | Included in Gemini Enterprise | Custom via Google Cloud | Per-seat |
| Amazon Q Developer | Free tier (limited) | Pro: $19/user/month | — | Custom via AWS | Per-seat |
| Cody (Sourcegraph) | Free (basic features) | Pro: $9/month | Enterprise: $19/user/month | Custom | Per-seat |
| Tabnine | Basic completions (limited) | Dev: $12/month (annual only) | — | Enterprise: $39/user/month | Per-seat (annual commitment required) |
Tool-by-Tool Pricing Breakdown
Cursor
Cursor offers four individual tiers. The Free (Hobby) plan gives you 2,000 completions per month and 50 slow premium requests — enough to evaluate the tool but not for daily professional use. Pro at $20/month is where most developers land: you get unlimited Tab completions, unlimited Auto mode (where Cursor chooses the model automatically), and a $20 credit pool for manually selecting frontier models like Claude Opus 4.6 or GPT-5.4. Pro+ at $60/month gives approximately 3× the usage. Ultra at $200/month provides approximately 20× the Pro allowance and priority access to new features. The Teams plan at $40/user/month adds shared rules, centralised billing, usage analytics, and SSO. The critical thing to understand: Auto mode is unlimited, but manually selecting premium models draws from your credit pool. Heavy Opus 4.6 usage can exhaust $20 in credits within a single intensive session. Monitor your credit dashboard weekly during your first month.
GitHub Copilot
Copilot’s five-tier structure makes it the most granular pricing in the market. Free gives you 2,000 code completions and 50 premium requests per month — enough for light, occasional use. Pro at $10/month is the standout value: unlimited completions, 300 premium requests, access to Claude Sonnet 4.6 and GPT-5.4, a coding agent, and code review features. Pro+ at $39/month bumps premium requests to 1,500 and unlocks every available model including Claude Opus 4.6. Business at $19/user/month adds centralised management, policy controls, and IP indemnification. Enterprise at $39/user/month includes knowledge bases, GitHub.com Chat, and custom model options — but requires GitHub Enterprise Cloud at $21/user/month, making the true cost $60/user/month. Overage beyond your premium request allocation costs $0.04 per request. Heavy agent users report needing 300–500 requests daily, which means Pro users can face significant overage charges.
Claude Code
Claude Code’s pricing works differently from IDE-based tools. It’s included with an Anthropic Claude Pro subscription at $20/month, which gives you a usage allowance for both Claude’s chat interface and Claude Code. Max 5× at $100/month multiplies the allowance fivefold. Max 20× at $200/month provides the heaviest usage tier. Alternatively, developers can use Claude Code through API tokens with their own key, paying per-token: $3/$15 per million tokens for Sonnet 4.6 or $15/$75 for Opus 4.6. API usage is the most flexible but least predictable — a single complex debugging session with Opus can consume 500K+ tokens, costing $15+ in one sitting. Average daily API costs for active Claude Code users run around $6 per developer, but can spike far higher during intensive agentic workflows.
Windsurf
Windsurf overhauled its pricing on March 19, 2026, replacing credits with daily/weekly quotas. The previous $15/month Pro price has risen to $20/month, eliminating its former cost advantage over Cursor. Free gives you a daily quota with unlimited Tab completions. Pro at $20/month provides a larger daily and weekly quota with access to frontier models (Claude Sonnet 4.6, GPT-5.4, Gemini 3.1 Pro). Teams at $40/user/month adds organisational features. Max at $200/month targets power users who hit daily limits on Pro. The quota system means your usage resets daily and weekly — you cannot sprint through your entire monthly allocation on one big project. Existing subscribers from before March 19 are grandfathered at old prices with a free extra week to evaluate the new system. Tab completions never consume quota on any plan.
Gemini Code Assist
Google’s AI coding assistant is free for individual developers — the most generous free tier in the category, with access to Gemini 3.1 Pro for code completion, chat, and code review. The Standard tier at $19/user/month adds enterprise features. Enterprise pricing is custom through Google Cloud. For developers already on Google Cloud, the integration with BigQuery, Cloud Run, Firebase, and Android Studio makes this effectively a free add-on to their existing infrastructure spend.
Amazon Q Developer
Amazon’s offering has a free tier with limited capabilities and a Pro tier at $19/user/month with full features including security scanning and AWS-specific integrations. Enterprise pricing is custom through AWS. The value proposition is narrow but deep: if your infrastructure runs on AWS, Q Developer understands your services at a level no general-purpose tool matches. Outside the AWS ecosystem, it offers little advantage.
Cody (Sourcegraph)
Cody’s pricing is the simplest in the market. Free provides basic code intelligence features. Pro at $9/month adds enhanced AI features with access to premium models. Enterprise at $19/user/month includes advanced codebase indexing, admin controls, and support. The $9/month price point makes it an affordable supplementary tool alongside a primary AI IDE.
Tabnine
Tabnine’s unique selling point is on-premises deployment, and its pricing reflects that enterprise positioning. A limited free tier offers basic completions. Dev at $12/month requires an annual commitment (no monthly option). Enterprise at $39/user/month includes on-premises deployment, custom model training, and advanced admin controls. For teams that cannot send code to external servers — regulated industries, government, defence — Tabnine is one of the few options at any price.
Hidden Costs to Watch
Credit and quota overages are the biggest surprise. Cursor’s $20 monthly credit pool sounds generous until you manually select Claude Opus 4.6 for a few complex agent sessions — frontier models burn credits at API rates, and a single heavy day can exhaust the entire pool. Similarly, Copilot’s $0.04/request overage seems trivial until agent mode racks up hundreds of requests in one session. Multiple developers report that heavy Copilot Agent usage can cost $100+ per month beyond the base subscription.
Model access restrictions on lower tiers often go unnoticed. Copilot Free limits you to lighter models — no Opus 4.6, no o3. Cursor’s Free tier throttles premium requests to “slow” speed, making agent work frustrating. Windsurf’s quota system means cheaper models consume less quota, so selecting a frontier model effectively reduces your daily usage cap.
Enterprise minimum seat requirements add friction. GitHub Enterprise Cloud (required for Copilot Enterprise) costs $21/user/month before you add Copilot, and typically requires minimum commitments. Tabnine Enterprise generally requires multi-year contracts for on-premises deployment.
Annual vs monthly billing creates meaningful differences. Cursor Pro drops from $20 to $16/month on annual plans (20% savings). Copilot Pro drops from $10 to $8.33/month ($100/year). Tabnine requires annual commitment entirely — there is no monthly option. The savings are real, but the AI coding tool landscape evolves so quickly that locking into an annual plan carries its own risk: the best tool in March may not be the best tool in September.
Best Value by Scenario
| Your Situation | Best Choice | Monthly Cost | Why |
|---|---|---|---|
| Solo developer on a budget | GitHub Copilot Pro | $10/month | Best feature-to-cost ratio in the market: unlimited completions, 300 premium requests, coding agent, multi-model access |
| Freelancer who needs flexibility | Cursor Pro + Copilot Free | $20/month | Cursor for complex client projects, Copilot Free for quick tasks — total cost $20 |
| Small team (5–10 devs) | GitHub Copilot Business | $19/user/month | IP indemnification, centralised management, and policy controls at under half the cost of Cursor Teams |
| Enterprise (50+ devs) | GitHub Copilot Enterprise + Cursor for power users | $39–$60/user/month | Copilot for the organisation-wide baseline; Cursor licences for developers doing complex agentic work |
| Student or learner | GitHub Copilot Free + Gemini Code Assist Free | $0 | Copilot Free gives 2,000 completions; Gemini Code Assist is fully free for individuals. Students also get Copilot Pro free via the GitHub Student Developer Pack |
| Privacy-critical / regulated | Tabnine Enterprise | $39/user/month | The only major tool offering fully on-premises, air-gapped deployment |
| Terminal-first power user | Claude Code (Max 5×) | $100/month | Largest context window, strongest agentic capabilities, parallel Agent Teams |
Price History and Trends
AI coding tool pricing has been volatile. In June 2025, Cursor replaced its fixed 500-request system with credit-based billing, effectively reducing monthly usage for the same price and prompting significant developer backlash. In March 2026, Windsurf eliminated credits entirely, switching to daily and weekly quotas and raising Pro from $15 to $20 — eliminating its former price advantage over Cursor. GitHub introduced premium request metering in mid-2025, making agent and chat usage a metered cost for the first time.
The trend is clear: headline prices are converging toward $20/month for individuals and $40/user/month for teams, but the real cost depends increasingly on your usage pattern and model selection. Providers are shifting from flat-rate to consumption-based models as frontier AI inference remains expensive.
Looking ahead, expect further convergence at the $20 standard tier, continued experimentation with billing mechanics (quotas, credits, requests), and potentially new “pay-per-outcome” models where you pay for completed tasks rather than raw usage. Open-source alternatives like Aider and OpenCode will continue to offer a free floor that keeps commercial pricing honest.
Frequently Asked Questions
Are free tiers good enough for real work?
For light use, yes. GitHub Copilot Free (2,000 completions + 50 premium requests) and Gemini Code Assist Free (individual) both provide genuine value for developers coding a few hours per week. Windsurf’s free tier is usable for occasional tasks. However, any developer coding more than two to three hours daily will hit limits quickly — premium models, agent mode, and chat features all consume restricted quotas on free plans. The free tiers are best understood as permanent trial versions, not production-grade tools.
Which tool gives the most for $20/month?
Cursor Pro — if you stay within Auto mode (unlimited) and use the $20 credit pool judiciously. You get the best AI-native IDE experience with the fastest autocomplete, powerful Composer agent, and access to every frontier model. The risk is overage if you manually select expensive models. For more predictable $20/month spend, Claude Code via Claude Pro gives you terminal-based agentic coding with the largest context window, but no IDE features. At $10/month (half the budget), Copilot Pro is arguably the smartest allocation, leaving $10 for a supplementary tool.
Do enterprise plans require annual contracts?
It varies. GitHub Copilot Business and Enterprise can be billed monthly, though multi-year commitments yield discounted rates. Tabnine requires annual contracts for all paid plans. Cursor Teams can be billed monthly or annually (20% discount for annual). Windsurf Teams is monthly with no forced annual commitment. For enterprise procurement, expect negotiated rates on multi-year, multi-seat agreements across all providers.
Read next:
- Best AI Coding Assistants in 2026: The Complete Comparison
- Cursor vs GitHub Copilot in 2026: Which Should You Choose?
- Is Cursor Worth It? An Honest Review After 6 Months
- Best Free AI Coding Tools in 2026
AI Agent Brief is editorially independent. Our recommendations are based on hands-on testing, not advertising relationships. When you subscribe to a tool through our links, we may earn a commission at no extra cost to you. This never influences our rankings.
© 2026 AI Agent Brief. All rights reserved.
Also in this series