AI Coding Tools Pricing: The June 2026 Reality Check

The spreadsheet you bookmarked six months ago is wrong. AI coding tool pricing has moved faster in the first half of 2026 than in any comparable period - new model tiers, renamed products, and at least one hard deadline that will change your monthly bill if you are on a claude.ai plan. Before you renew anything or add a seat for a new hire, check the numbers below.

The short answer for solo developers in June 2026: budget $20 to $40 per month for a serious AI coding setup. Small teams of 5 to 10 people should expect $200 to $500 per month in tooling costs, before any usage-based overages from agentic workflows. Enterprise buyers deal with per-seat contracts, but the floor is now around $39 to $100 per seat depending on model access requirements.

Last updated: June 10, 2026

The June 2026 Deadline You Cannot Miss

Claude Fable 5 access on claude.ai Pro and Max plans ends on June 22, 2026. After that date, Fable 5 becomes a paid API tier only, priced at $10 per million input tokens and $50 per million output tokens. If your team relies on Fable 5 through the claude.ai interface rather than the API, you need a plan before that date. A heavy day of agentic coding work can consume enough context to make the per-token math sting quickly - see our Fable 5 production cost modeling piece for concrete estimates.

The API comparison: Fable 5 at $10/$50 per MTok versus Opus 4.8 at $5/$25 per MTok. Fable 5 costs exactly 2x Opus 4.8 at the API level. For long-running agents where the model quality genuinely changes task completion rates, that may be worth it. For routine autocomplete and refactors, it probably is not.

Master Pricing Table - June 2026

Note: GitHub Copilot new plan sign-ups for Pro, Pro+, and Max are temporarily paused as of this writing. Existing customers can upgrade between tiers.

Per-Tool Breakdown

Claude Code and Claude Plans

Anthropic now publishes pricing at claude.com/pricing (note: not anthropic.com/pricing, which also exists and covers the same data). The tier structure as of June 10, 2026:

Free - $0. Standard Claude access with usage limits. No credit card required.

Pro - $17/month billed annually ($200 upfront) or $20/month billed monthly. This is the baseline serious plan and includes Claude Code access.

Max - starts at $100/month. Anthropic offers two Max variants: Max 5x (5 times the usage of Pro) and Max 20x (20 times Pro usage). The "from $100" framing on the pricing page signals the 5x tier sits at $100, with 20x priced higher - specific Max 20x pricing is shown after you select the tier on the signup flow.

Team - $20 per seat per month billed annually, $25 per seat billed monthly. Includes a "Premium seat" option at $100/seat (annual) / $125/seat (monthly) which provides 5x more usage than standard seats plus Claude Code and Cowork Connect (Microsoft 365 and Slack integrations).

Claude Code via API - $0.08 per session-hour for active runtime, on top of the underlying model token costs.

API token pricing (current as of June 10, 2026):

Prompt caching write costs 25% more than input; cache reads cost about 10% of input price.

Cursor

Cursor's JSON-structured pricing (extracted from their pricing page on June 10, 2026):

Hobby - $0. Limited Agent requests, limited Tab completions. No credit card required.

Pro - $20/month. Extended agent limits, frontier model access, MCPs, skills, hooks, cloud agents. This is where most individual developers land.

Pro+ - $60/month. Higher limits than Pro, same feature set.

Ultra - $200/month. Maximum quotas, same features as Pro+ with the highest compute ceiling.

Teams - $40/user/month (Standard tier). A Premium Teams tier also exists - listed as "Everything on Individual, plus centralized billing, team marketplace for internal rules and skills, agentic code reviews."

What changed since Q1: Cursor added Pro+ as a distinct tier between Pro and Ultra, giving users a step before committing to $200/month. The Teams plan previously sat at a different price point; the current $40/user reflects a mid-2026 update.

GitHub Copilot

GitHub's individual plans as of June 10, 2026:

Free - $0. 2,000 code completions per month, access to Haiku 4.5 and GPT-5 mini, Copilot CLI. No credit card required.

Pro - $10/month. Unlimited code completions, cloud agent access, code review agent, access to third-party agents including Claude Code and OpenAI Codex, $15/month in included AI credits.

Pro+ - $39/month. Everything in Pro plus premium model access including Opus, audit logs, and 4x+ the included usage versus Pro. $70/month in included credits.

Max - $100/month. Priority access to new models, 2.9x+ more usage than Pro+. $200/month in included AI credits.

Business plans are separate and start at $19/user/month. Enterprise pricing requires a quote.

Important caveat: new sign-ups for Pro, Pro+, and Max are temporarily paused as GitHub manages capacity. This is worth watching if you are evaluating Copilot for new seats.

Windsurf (Now Devin Desktop)

Windsurf was acquired and rebranded. The windsurf.com domain now redirects to Devin (by Cognition), and the IDE is called "Devin Desktop." Pricing as of June 10, 2026 from devin.ai/pricing:

Free - $0. Light quota for agentic coding, limited model availability, unlimited inline edits, unlimited Tab completions.

Pro - $20/month. Increased quotas including OpenAI, Claude, and Gemini frontier models, full model availability, SWE 1.6 (their current flagship model), access to Devin Cloud (cloud agents). Additional usage available at API pricing.

Max - $200/month. Significantly higher quotas than Pro, otherwise same feature set.

Teams - $80/month for the base team plan plus $40/month per full developer seat. Includes centralized billing, collaboration features, admin dashboard, priority support, and unlimited team members with flex seat pricing.

Enterprise - Contact for pricing. Adds VPC deployment, SAML/OIDC SSO, dedicated account management.

Hidden Costs: What the Pricing Pages Do Not Lead With

Usage caps and overage. Every tool has a "soft" usage limit where you hit degraded service or get prompted to upgrade before you see hard overages. For Cursor, Agent request limits are the primary pressure point. For Copilot plans, the dollar credit amounts ($15/$70/$200 per month) are the real ceiling - requests beyond included credits consume additional billing.

Agentic token burn. Running an AI agent on a multi-file refactor or a debug session with tool calls can consume 50,000 to 200,000 tokens per task, depending on context window and model. At Fable 5 rates ($10/$50 per MTok), a single complex agentic session can cost $2 to $8. Run 20 such sessions in a day through the API and you are at $40 to $160 before any subscription cost. This math is why claude.ai Max plans exist - they pre-purchase a ceiling.

Model selection risk. Tools that give you model choice (Cursor, Copilot Pro+, Devin Pro) can inadvertently run expensive models on routine tasks. Cursor defaults to a sensible model for completions but agents may escalate. Copilot's credit system means selecting Opus for a quick question eats more of your $70 monthly credit than selecting Sonnet.

The Fable 5 overhang. Until June 22, 2026, Fable 5 is accessible on claude.ai Pro and Max plans within their usage limits. After that date, API-level access only. Teams that built workflows around Fable 5 quality expectations need to decide: move to the API (with explicit cost modeling per task) or downgrade to Opus 4.8 (half the API price, meaningfully different quality on complex reasoning).

Seat minimums. Copilot Business requires a minimum number of seats. Devin Teams has a $80/month base before you add any users. These floors matter for 1-3 person teams where per-seat tools may be cheaper individually.

Recommendations by Persona

Solo developer, $20-30/month budget: Cursor Pro ($20) gives you the most capable IDE-native experience with frontier models. Alternatively, Claude.ai Pro ($17-20) with Claude Code covers both chat and terminal-native workflows. Copilot Free is a viable third option if you primarily want completions rather than agents.

Solo developer wanting maximum output: Cursor Ultra at $200/month or Claude Max at $100/month. If you are doing heavy agentic work daily, Max 5x ($100) plus API access for Fable 5 where quality matters is a reasonable setup for under $200/month total.

Small team (3-10 developers): Claude Team at $20/seat/month (annual) is the lowest per-seat cost for a full-featured AI coding environment. Copilot Pro at $10/seat/month is lower if completions are the primary use case and agents are secondary. Windsurf/Devin Teams at $80 base + $40/seat is only competitive above 5 seats.

Enterprise (50+ seats, compliance requirements): GitHub Copilot Enterprise or Anthropic Team/Enterprise - both offer SSO, audit logs, and usage controls. Copilot Enterprise pricing requires a quote. Anthropic Enterprise is contact-based. Devin Enterprise adds VPC deployment for teams with strict data residency needs.

API-first / product builders: Fable 5 at $10/$50 per MTok for tasks that genuinely need top-tier intelligence. Sonnet 4.5 at $3/$15 for the bulk of production tasks. Haiku at $1/$5 for high-volume, low-complexity work. The Fable 5 cost-per-task analysis covers how to model the break-even between Fable 5 and Opus 4.8 for your specific workload.

---

FAQ

How much does Claude Code cost in 2026?

Claude Code is included in Claude Pro ($17-20/month), Max ($100+/month), and Team ($20/seat/month on annual billing) plans. When accessed via the API rather than claude.ai, Claude Code charges $0.08 per session-hour for active runtime on top of standard model token costs. For heavy agentic use, Max 5x at $100/month is the practical floor to avoid hitting usage caps.

What is the cheapest AI coding tool with a free tier in 2026?

GitHub Copilot Free includes 2,000 code completions per month and basic Copilot CLI access at no cost. Cursor Hobby and Windsurf/Devin Free both offer unlimited Tab completions with limited agent quota. All three require no credit card. For solo projects and light use, Copilot Free or Cursor Hobby covers basic needs.

How does Cursor pricing compare to GitHub Copilot in 2026?

At the individual level: Cursor Pro ($20/month) and Copilot Pro ($10/month) are the direct comparison. Cursor includes more agentic capability and MCP/skills support; Copilot Pro integrates more tightly with GitHub workflows and PR-level agents. Copilot Pro+ ($39/month) is closer to Cursor Pro in model quality. For teams, Cursor is $40/user/month vs Copilot Business at $19/user/month, though feature sets differ.

What changed with Windsurf pricing after the Cognition acquisition?

Windsurf was rebranded as Devin Desktop following its acquisition by Cognition. The windsurf.com domain redirects to devin.ai. Pricing structure is now: Free, Pro ($20/month), Max ($200/month), Teams ($80/month base + $40/seat). The underlying model is SWE 1.6 rather than the prior Windsurf-specific model. Existing Windsurf subscribers may see different terms during transition.

Is Fable 5 free on claude.ai plans after June 22, 2026?

No. Anthropic has confirmed that Fable 5 access through claude.ai (Pro, Max, Team plans) is available only until June 22, 2026. After that date, Fable 5 access requires using the Anthropic API directly, billed at $10 per million input tokens and $50 per million output tokens. The models remaining on claude.ai plans after June 22 include Opus 4.8 and Sonnet 4.5. See our June 22 deadline breakdown for transition guidance.

---

Sources

• Anthropic / Claude pricing - claude.com/pricing (anthropic.com/pricing redirects to same) - checked June 10, 2026
• Cursor pricing - cursor.com/pricing - checked June 10, 2026
• GitHub Copilot plans - github.com/features/copilot/plans - checked June 10, 2026
• Windsurf / Devin pricing - devin.ai/pricing (windsurf.com/pricing redirects here) - checked June 10, 2026
• OpenAI pricing - openai.com/pricing - page requires JavaScript to render; individual plan pricing cited from OpenAI API documentation as of last verified data; API pricing current as of June 2026