OpenAI Codex

OpenAI Codex is the coding agent from OpenAI for building and shipping software autonomously, powered by GPT-5.5 and accessed through ChatGPT, the Codex desktop app for macOS and Windows, a VS Code extension (also compatible with Cursor and Windsurf), JetBrains IDE integration, and a CLI via npm. Codex runs multi-step development tasks end to end: writing features, fixing bugs, refactoring, running tests, and filing PRs for human review, all in isolated cloud sandboxes preloaded with GitHub repositories. The platform includes four core agentic systems. Multi-agent workflows run parallel agents simultaneously in isolated git worktrees. Skills let Codex go beyond writing code to directly contribute to code understanding, prototyping, and documentation aligned with team standards. Automations enable Codex to work unprompted on routine but important tasks including issue triage, alert monitoring, and CI/CD, so teams stay focused on building. Sites, launched June 2, 2026, lets users create, deploy, and manage hosted web projects directly within the Codex app, along with Annotations for in-place code editing and domain-specific workflows for enterprise roles beyond engineering. Integration depth is exceptional across IDE, SCM, and collaboration surfaces. Codex ships an official VS Code extension (compatible with Cursor and Windsurf), native JetBrains integration (IntelliJ, PyCharm, WebStorm, Rider), and Xcode support via Apple's Xcode 26.3. SCM connects natively to GitHub and GitLab. The platform includes 60+ app integrations including Slack, Google Drive, SharePoint, and Atlassian. MCP is supported. A GitHub Action enables CI/CD automation. Subagents allow Codex to spawn parallel workers for complex tasks. Codex with GPT-5.5 is now generally available on Amazon Bedrock. Bitbucket is not supported for SCM; only GitHub and GitLab repositories can be connected. Reference customers include Ramp, Duolingo, Harvey, Sierra, Cisco Meraki, and NVIDIA. Pricing runs through ChatGPT plans. Free includes limited Codex access. Go at approximately $8/month provides more messages and access. Plus at $20/month (billed monthly) provides expanded Codex usage, advanced reasoning with GPT-5.5 Thinking, and expanded deep research and agent mode. Pro at $200/month provides the highest Codex access with priority to frontier models including GPT-5.5 Pro. Business at $25/user/month (annual) includes everything in Plus plus 60+ app integrations, SAML SSO, MFA, and no training on customer data. Enterprise is custom-priced with additional security, compliance controls, data residency in ten regions (US, EU, UK, JP, CA, KR, SG, IN, AU, UAE), SCIM, Enterprise Key Management, IP allowlisting, and 24/7 priority support with dedicated onboarding. Annual billing applies on Business and Enterprise tiers. Codex is not the right fit for organizations that require fully self-hosted or on-premises AI tooling, as all Codex tasks run in OpenAI cloud sandboxes with no local deployment option. Teams needing Bitbucket SCM support will not find it here. The credit and rate-limit structure varies by plan tier and can throttle heavy users on Plus during peak periods. Teams that prefer local-first agentic coding should evaluate Cursor at $20/month or Claude Code at $20/month instead. Current state Q2 2026: GPT-5.5 launched April 23 and is now the default model powering Codex, succeeding GPT-5.4 (March 2026) and GPT-5.3-Codex (February 2026). GPT-5.2 and GPT-5.3-Codex are being sunset with API discontinuation on June 30, 2026. Sites, Annotations, and domain-specific enterprise workflows launched June 2. Codex and GPT-5.5 became generally available on Amazon Bedrock. Codex is SOC 2 Type II and ISO 27001 certified. Customer data is not used to train models on paid tiers.