Coding agents
Gartner's first Magic Quadrant for coding agents
Gartner's first Magic Quadrant for coding agents names Anthropic, Cursor, GitHub and OpenAI Leaders, and expects the IDE to become optional for most teams.
On 20 May 2026, Gartner published the first edition of its Magic Quadrant for Enterprise AI Coding Agents. Twelve vendors are evaluated. Four are named Leaders: Anthropic (Claude Code), Cursor, GitHub (Copilot) and OpenAI (Codex). AWS, Google, Alibaba Cloud (with Qoder) and Cognition are Challengers. Tabnine is the only Visionary; Atlassian, BytePlus and JetBrains are Niche Players (Virtualization Review).
A new category, not just a new name
The 2024 and 2025 editions covered "AI Code Assistants". Gartner now uses a stricter definition: an enterprise coding agent perceives context, turns intent into a multistep plan, then executes and verifies those steps across code, tests and related artefacts. Suggesting a line or answering in a chat is no longer enough.
The visible consequence: AWS and Google, Leaders in earlier editions, drop to Challengers. GitHub stays a Leader and is placed highest for ability to execute (GitHub); Cursor is furthest on completeness of vision (Cursor). Gartner also describes model providers "moving up the stack": Anthropic and OpenAI now sell their own agents, competing directly with tools built on their APIs.
The Leaders' figures
These figures come from the vendors themselves, not from Gartner:
- GitHub Copilot: more than 4.7 million paid subscribers at the end of January 2026, up 75% year over year (Microsoft), and 140,000 customer organisations in May, according to GitHub.
- Cursor: more than 50,000 customer teams, including nearly 70% of the Fortune 1000 (Cursor, February 2026).
- Codex: more than 4 million weekly users (OpenAI).
The IDE is no longer the centre
The report comes with planning assumptions reserved for clients. Gartner made one public: by 2027, more than 65% of engineering teams using agentic coding will treat the IDE as optional, with control, governance and validation moving to automated platforms (Gartner press release). On productivity, Gartner says 90% of engineering leaders report improvements, with a net average productivity gain of 19.3% (Gartner).
Gartner sizes the market at $9.8 billion to $11 billion annualised as of April 2026. In a June 24, 2026 press release, it also warns that by 2028, AI coding costs will exceed the average developer's salary, driven by token consumption and usage-based licences (Gartner).
What it changes for developers
- MCP becomes mandatory: native Model Context Protocol support is on the list of required capabilities. An internal tool without an MCP server will be harder to plug into an agent.
- Human review is still required: approval of agent-produced changes, detailed logs of agent actions, and a guarantee that the model is not trained on customer code.
- Billing is changing: pricing moves from seats to usage (tasks, premium requests, tokens). Consumption needs to be tracked like any other budget line.
To choose between the three most-cited agents, see our Claude Code, Codex and Copilot comparison.
The limits
The full report is paywalled; this article relies on public excerpts shared by the press and by the ranked vendors. Gartner itself states that its research reflects its analysts' opinions. It also writes that without high-quality context, strict controls and automated validation, agents increase rework, security risk and architectural drift. Independent measurements remain cautious: see our piece on the security of AI-generated code.
These are planning assumptions, not certain predictions: read them as a direction, not a date.
Sources: Gartner, Magic Quadrant for Enterprise AI Coding Agents, 20 May 2026; Virtualization Review, 5 June 2026; Microsoft, FY26 Q2 earnings call; GitHub, Cursor and OpenAI.
Spotted an error in this article? Report it · Our corrections policy
