HarnessMatch

Search 73 product records by name, workflow, interface, or documented capability.

Quick access
Use ↑ ↓ to move, Enter to open, Esc to close.

Quick access links shown.

Menu

Coding agent

Codex

Agentic coding across app, terminal, IDE, cloud, and automation.

Status
Active
License
Apache-2.0 client
Evidence
Documented + code-verifiable + independently measured configuration, 49 sources
Product record checked
2026-08-22
Compare harnessesInspect sources

At a glance

OpenAI's coding agent across desktop, terminal, IDE, cloud, SDK, and CI, with default local sandboxing, managed permission policy, MCP, hooks, subagents, persistent memories, worktrees, browser tools, and optional local or custom model providers.

Good choice if

  • ChatGPT subscribers who want one agent across app, IDE, and terminal
  • Headless repository tasks and CI-style workflows
  • Users who value sandbox boundaries and parallel delegation

Check before choosing

  • The standard local workspace sandbox limits writes and disables network access, but legacy modes can still read beyond the project; least-privilege permission profiles are beta and do not compose with legacy sandbox settings
  • Auto-review can replace a human at eligible sandbox-boundary prompts, including through the --approve-for-me CLI flag, but it only reviews requested escalations, can make mistakes, and is not a deterministic security guarantee
  • Worktrees isolate Git checkouts rather than processes and are a desktop-app workflow; local chats can still operate directly in the user's checkout
See 6 more considerations
  • There is no product-level file checkpoint: review-pane Git revert and session resume do not reverse arbitrary shell, browser, MCP, or remote-system side effects
  • Browser and Computer Use can act outside the repository; the browser uses a separate profile, while Computer Use requires OS-level screen and accessibility access plus attended approval
  • Remote connections inherit the connected host's projects, files, credentials, permissions, plugins, browser setup, and local tools rather than creating a separate isolation boundary
  • Hosted access remains centered on OpenAI models; custom, Bedrock, Ollama, and LM Studio paths add flexibility but model quality and compatibility are separate from the harness
  • Codex Security is an optional plugin or cloud vulnerability workflow, not evidence that the base harness or selected model will find every vulnerability
  • The Apache-2.0 repository exposes the CLI, SDK, app server, sandbox, tests, and rollout tooling, but not the proprietary IDE extension or Codex cloud service

Capability support

Documented first-class product support, checked against the sources below.

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Reusable skills
DocumentedProduct-supported reusable skill packagesSource · checked 2026-07-30Support does not establish portability, package quality, safety, or adoption.
Local models
DocumentedLocal or self-hosted model pathSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
DocumentedNon-interactive or automation surfaceSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Browser control
DocumentedBuilt-in or product-supported browser controlSource · checked 2026-07-30The source establishes the mechanism, not its quality or availability in every mode.
Isolated execution
Available by defaultSandbox-first local CLI and managed cloud executionSource · checked 2026-07-30The effective boundary still depends on the selected sandbox and approval policy.
Undo file changes
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.

Getting started

Install the Codex CLI or extension, authenticate, then run `codex` interactively or `codex exec` for automation.

Open official documentation

Classification and operating model

Category fit and technical mechanisms are evidence records, not product-quality scores.

Category fit
Qualifies, 4/4 criteria
Operating model
7/7 layers documented
Inspect category criteria and operating mechanismsFirst-party records

Why it qualifies as a coding harness

This confirms category fit, not product quality. Every required criterion links back to first-party evidence.

Qualifies4 of 4 required criteria evidenced

  • Adaptive agent loop

    Documented

    The system repeatedly observes results and chooses the next action instead of following a fixed one-pass graph.

    EvidenceCodex CLI

  • Repository tool execution

    Documented

    The system can use tools to inspect and change a repository or its execution environment.

    EvidenceDeveloper command reference

  • Task-aware context management

    Documented

    The runtime assembles, updates, compacts, retrieves, or persists task-relevant context while work proceeds.

    EvidenceAGENTS.md instructions

  • Model-independent runtime control

    Documented

    Permissions, budgets, interruption, policy, or stop controls operate outside the model's own text generation.

    EvidenceApprovals and security

Membership establishes category fit only. It does not score quality, safety, autonomy, model capability, or benchmark performance. · Read the membership rule.

How it works under the hood

Seven mechanisms mapped from first-party records. These labels describe what the harness provides, not how intelligent its model is.

7/7layers documented
  • Execution & isolationSandbox availableDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible + browserDocumented mechanism, not a performance score.
  • Context & statePersistent stateDocumented mechanism, not a performance score.
  • Lifecycle & recoverySession resumeDocumented mechanism, not a performance score.
  • ObservabilityStructured tracesDocumented mechanism, not a performance score.
  • VerificationTool-assistedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Measured and public context

Configuration-specific measurements and source-native activity stay separate from general product capability.

Inspect code audit, measured configurations, and ecosystem signalsContext, not a product score

Public code audit

4/5public artifacts present
Security policy
Present at inspected commit
CI workflow
Present at inspected commit
Automated tests
Present at inspected commit
Evaluation assets
Not found
Contributor documentation
Present at inspected commit

The stable 0.145.0 tree exposes the CLI, SDK, app server, OS sandboxes, rollout trace implementation, 620 test-like files, and 26 workflows. It does not expose the proprietary IDE extension or Codex cloud service, and no product benchmark suite was found.

Inspect commit 25af12f7e61572b0bc18ddb1008be543b91519b0, checked 2026-07-27
Context, not quality

Public ecosystem signals

Source-native observations for exact mapped artifacts and reviewed stable release trains. Different units and populations stay separate, and missing coverage is never treated as zero.

View this harness in Usage
  • Latest stable releaserust-v0.153.4Released 2026-09-04; 33 stable releases in 90 daysOpen release
  • OpenRouter 30d tokens3.16T#8 coding app; 2026-08-09 to 2026-09-07Open app page
  • Homebrew 30d events132.22KCask: codexOpen artifact
  • npm last-month downloads76.02MPackage: @openai/codexOpen artifact
  • Release asset downloads9.83M157 stable releases; 2604 matched assetsOpen artifact
  • VS Code installs14.01MExtension: openai.chatgptOpen artifact
  • Open VSX downloads11.54MExtension: openai/chatgpt; latest 26.5901.22334Open artifact
  • GitHub stars122.39KClient-source repository; 18.81K forksOpen artifact

Routing, package retrievals, release downloads, editor installs, and repository interest observe different populations. They are never added together and never affect capability evidence, classification, or measured results.

Interpretation rulesSignals checked

First-party evidence

Each capability claim links to the first-party record that supports it.

49 first-party sourcesProduct record checked

Product and interfaces

8 sources
View 7 more sources

Execution and control

9 sources
View 8 more sources

Agents, state and recovery

9 sources
View 8 more sources

Automation and extensions

12 sources
View 11 more sources

Enterprise and operations

5 sources
View 4 more sources

Releases and public code audit

6 sources
View 5 more sources