GitHub Copilot, Cursor, Claude Code and Gemini Code Assist: plans compared

Choosing an AI coding assistant used to mean comparing autocomplete quality. In 2026 the products have converged on agents that read a repository, edit several files, run commands and open pull requests, and the differences that matter to a small team are now where the tool runs, how usage is billed, and what administrative controls come with each plan.

This comparison covers four widely used options: GitHub Copilot, Cursor, Claude Code and Gemini Code Assist. It is based on each vendor's own pricing and documentation pages as published on 7 October 2026. We have not benchmarked the tools against each other; output quality depends heavily on your codebase and the model each tool uses, so the last section explains how to run your own short trial instead.

The short version

The Gemini figures are our conversion of Google's hourly licence fees at 730 hours a month; Google bills the hourly rates shown on its pricing page.

GitHub Copilot

Copilot's individual plans are Free, Pro ($10 a month), Pro+ ($39) and Max ($100). Organisations choose Copilot Business at $19 per granted seat per month or Copilot Enterprise at $39. Verified students get Copilot Student free.

Usage is now metered in GitHub AI Credits. GitHub's plans documentation lists monthly allowances of 1,500 credits for Pro (1,000 base plus a 500 flex allotment), 7,000 for Pro+, 20,000 for Max, 1,900 per user for Business and 3,900 per user for Enterprise. The Free plan includes limited chat and agent usage and 2,000 inline suggestions a month, but no cloud agent and only "review selection" for code review in VS Code.

From Pro upwards you get agent mode, Copilot code review, the cloud agent that can be assigned work and open pull requests, the CLI and custom instructions. Model choice scales with the plan: Free and Student use automatic model selection only, Pro offers a selection of models, and Pro+, Max and Enterprise get priority access to premium models. Pro+ adds delegation to third-party agents such as Claude and OpenAI Codex (in preview). Enterprise adds GitHub.com integration and codebase indexing.

Fits best when your code already lives on GitHub and you want one licence that covers the editor, the terminal and pull requests.

Cursor

Cursor is an editor built around AI. Its Hobby plan is free with limited agent requests. Individual costs $20 a month and adds extended agent limits, access to frontier models, MCPs, skills and hooks, and cloud agents, with Bugbot code review on usage-based billing. Teams costs $40 per user per month and adds centralised billing, a team marketplace for rules and plugins, shared cloud agents, Bugbot reviews, usage analytics, team-wide privacy mode and SAML/OIDC single sign-on. Enterprise is custom-priced with pooled usage, SCIM, repository, model and MCP access controls, audit logs and an AI code tracking API.

Fits best when developers are willing to switch editor and want the AI features designed into the editing experience. Note that SSO starts at the Teams tier.

Claude Code

Claude Code is Anthropic's agentic coding tool. Its documentation describes it as reading your codebase, editing files, running commands and integrating with development tools, available in the terminal, VS Code (and Cursor), JetBrains IDEs, a desktop app and the browser at claude.ai/code. It also runs in CI through GitHub Actions or GitLab CI/CD, can review pull requests, and connects to other systems through MCP. A project-level CLAUDE.md file holds team conventions, and it can read an existing AGENTS.md instead.

It is included in Claude's paid plans: Pro at $20 a month billed monthly ($17 a month with annual billing), Max from $100 a month, Team (standard seat $25 a month billed monthly or $20 annually; premium seat $125 or $100), and Enterprise at $20 per seat per month billed annually plus usage at API rates. It is not included in the Free plan. The terminal CLI, VS Code and JetBrains integrations can also be used with an Anthropic Console (API) account or supported third-party cloud providers, which suits teams that prefer pay-per-token billing.

# macOS, Linux, WSL (native installer from the official docs)
curl -fsSL https://claude.ai/install.sh | bash
cd your-project
claude

Fits best when you want an agent that works the same way from the terminal, the IDE and CI, and you are comfortable reviewing its changes as diffs.

Gemini Code Assist

Google's developer documentation now describes Gemini Code Assist as available in two editions managed through Google Cloud: Standard and Enterprise. Both include agent mode and the Gemini CLI, an open-source agent for the terminal; Enterprise adds code customisation based on your private repositories, Gemini in Apigee and Application Integration, extra Gemini Cloud Assist features and higher daily agent limits. Pricing is published as hourly licence fees: Standard is $0.031232877 an hour on a monthly commitment or $0.026027397 on a 12-month commitment; Enterprise is $0.073972603 or $0.061643836. Subscriptions are billed monthly.

Google's pricing page points individuals to a separate free version, "Gemini Code Assist for individuals", at codeassist.google, while its developer documentation lists deprecation notices for the consumer version and consumer accounts. Check the current status there before standardising on the free tier.

Fits best when your team already runs on Google Cloud and wants licences managed through the same admin console.

Questions to ask before you buy

• Where do your developers work? An editor-centric tool and a terminal-centric agent suit different habits.
• How is heavy usage billed? Copilot meters credits, Cursor bills some features on usage, Claude Enterprise adds API-rate usage to the seat price. Ask what happens when a developer runs out mid-month and whether you can set budgets.
• What are the data terms? Read each vendor's privacy and data-retention terms for your plan tier; business tiers typically differ from individual ones. Cursor lists team-wide privacy mode from Teams upwards.
• Do you need SSO, SCIM and audit logs? These usually sit on team or enterprise tiers only.
• Can you limit what the agent may do? Look for controls over which commands run automatically, network access and which MCP servers are allowed.

How to run a fair two-week trial

1. Pick three to five real tasks from your backlog: a bug fix, a small feature, a test-writing task, a refactor and a documentation update.
2. Have two developers complete each task with each shortlisted tool, timing the work and recording how many agent suggestions were accepted unchanged.
3. Run your normal code review on every result. Count review comments and defects found later, not just speed.
4. Check the bill or credit usage at the end of the trial and project it to your full team.
5. Write down the tool's permission settings you used, so the eventual rollout uses the same guardrails.

Whichever tool you choose, treat AI-generated code like any other contribution: it goes through review, tests and security scanning before it merges.

Guardrails to configure on day one

Whichever assistant you choose, agents that can run commands and push code need the same basic guardrails you would give a new contractor:

• Branch protection. Require pull requests and at least one human approval on your main branch, so no agent can merge on its own.
• Command permissions. Start with the agent asking before it runs shell commands or edits files outside the project, then relax only the commands you trust (for example, running tests).
• Secrets out of reach. Keep production credentials out of developer machines and repositories; agents read what the developer can read.
• Approved MCP servers only. Every connected MCP server extends what the agent can do. Keep a short allowlist and review new servers like any other dependency.
• Project instructions. Put your conventions, test commands and "never do" rules in the instruction file the tool reads (for example CLAUDE.md or AGENTS.md, or custom instructions in Copilot), so every developer's agent behaves consistently.
• Budget alerts. Set spending limits wherever the vendor offers usage-based billing.

Summary

For a GitHub-centred team, Copilot Business at $19 per seat is among the lowest team prices listed here. Cursor suits developers who want an AI-first editor and are willing to switch. Claude Code fits teams who want the same agent in the terminal, IDE and CI and who may prefer API-based billing. Gemini Code Assist is the natural choice for organisations already managing licences in Google Cloud. Prices and allowances change frequently in this market; re-check the vendor pages linked below before you commit.