AccessiMind - AI WCAG Accessibility

AccessiMind logo

AccessiMind is a VS Code extension for finding and improving accessibility issues in code. It combines WCAG 2.2 analysis, AI-assisted fixes, provider model selection, agentic planning, browser inspection workflows, Jira-ready task generation, and usage statistics.

The public command namespace is still wcagEnhancer.* for backward compatibility, while user-facing branding is AccessiMind.

Current Capabilities

WCAG 2.2 Code Analysis

Analyze the active file or selected code from the editor context menu or Command Palette.
Apply AI-generated accessibility improvements directly to the editor.
Target WCAG levels A, AA, or AAA.
Use strict mode, context-aware analysis, custom rule files, and disability-focus targeting.
Prioritize screen reader, low vision, hearing, motor, or cognitive accessibility needs while keeping baseline WCAG coverage.

AI Providers

AccessiMind can use several provider families:

Provider	Use case	Requirement
Gemini	Google Gemini API models for general WCAG analysis and fixes	Gemini API key
VS Code Copilot	GitHub Copilot-backed language models through the VS Code Language Model API	Active Copilot entitlement
Ollama	Local/private model execution	Local Ollama server
Codex Subscription	Codex CLI through ChatGPT/Codex sign-in	Installed Codex CLI and signed-in ChatGPT/Codex account

Model selection is dynamic through AccessiMind: Select Provider Model. Gemini discovery uses the Google models API when an API key is available, Copilot discovery uses VS Code's language model API, Ollama discovery reads local /api/tags, and Codex Subscription runs through local codex exec.

Agent Session

AccessiMind: Agent Session starts a guided workflow that:

Uses the currently selected AI provider.
Collects active editor context, selected text, and a bounded workspace file map.
Creates a Markdown implementation plan before editing.
Generates code only after confirmation.
Opens a diff and applies changes only after explicit approval.

Built-in workflows include fixing the active file, fixing a selection, investigating the workspace, and creating a production plan.

Browser And Companion Extension

AccessiMind includes live browser-oriented workflows:

AccessiMind: Show in Browser creates a temporary HTML preview for the active file.
Browser integration settings control the live Chrome/Edge inspector bridge.
The companion browser extension lives in browser-extension/accessimind-companion/.
Browser bridge utilities support selection capture and live inspection flows.

Browser integration settings:

Setting	Purpose
`wcagEnhancer.browserIntegration.enabled`	Enables live browser inspection support
`wcagEnhancer.browserIntegration.browserPath`	Optional Chrome or Edge executable path
`wcagEnhancer.browserIntegration.launchMode`	Launch an isolated debugging browser or attach to an existing debugging session

Chat And ChatGPT Handoff

AccessiMind: Chat opens the in-sidebar chat view.
AccessiMind: Inline Chat edits selected code with natural-language instructions.
AccessiMind: Open in ChatGPT builds an accessibility prompt, copies the full prompt to the clipboard, and opens ChatGPT or a configured ChatGPT app URL.
AccessiMind: Configure ChatGPT Auth stores ChatGPT app and MCP server URLs for account-authorized handoff flows.
The ChatGPT bridge is a handoff path. It does not reuse hidden browser sessions or cookies.

Jira, Statistics, And Settings

Generate Jira-ready accessibility tasks from findings.
Export and reset statistics.
Track improvements by time period, language, provider, model, and WCAG criteria.
Export, import, restore, clear, and inspect persistent settings.
Sync supported settings with the AccessiMind JSON-backed settings file.

Commands

Command title	Command id	Description
AccessiMind: Analyze Open File	`wcagEnhancer.analyzeOpenCode`	Analyze and improve the active file
AccessiMind: Analyze Selection	`wcagEnhancer.analyzeSelectedCode`	Analyze and improve selected code
AccessiMind: Preview Improvement	`wcagEnhancer.previewImprovement`	Preview an AI improvement
AccessiMind: Chat	`wcagEnhancer.openChat`	Focus the AccessiMind chat view
AccessiMind: Inline Chat	`wcagEnhancer.inlineChat`	Modify selected code from an instruction
AccessiMind: Settings	`wcagEnhancer.openSettings`	Open the native settings UI
AccessiMind: Help & Documentation	`wcagEnhancer.openHelp`	Open the help panel
AccessiMind: Show in Browser	`wcagEnhancer.showInBrowser`	Open a browser preview for the active file
AccessiMind: Agent Session	`wcagEnhancer.startAgentSession`	Start guided plan, generate, diff, and apply flow
AccessiMind: Select Provider Model	`wcagEnhancer.selectProviderModel`	Refresh and select models for the active provider
AccessiMind: Connect Codex Account	`wcagEnhancer.connectCodexAccount`	Select Codex Subscription and open `codex login`
AccessiMind: Test Codex Account	`wcagEnhancer.testCodexAccount`	Validate the local Codex CLI account
AccessiMind: Open in ChatGPT	`wcagEnhancer.openChatGptBridge`	Copy prompt and open ChatGPT handoff
AccessiMind: Configure ChatGPT Auth	`wcagEnhancer.configureChatGptAuth`	Configure ChatGPT app and MCP URLs
AccessiMind: ChatGPT Auth Guide	`wcagEnhancer.openChatGptAuthGuide`	Open guidance for account-authorized ChatGPT handoff
AccessiMind: Create Jira Task	`wcagEnhancer.createJiraTask`	Generate a Jira-compatible accessibility task
AccessiMind: Statistics	`wcagEnhancer.showDetailedStatistics`	View detailed statistics
AccessiMind: Export Statistics	`wcagEnhancer.exportStatistics`	Export statistics
AccessiMind: Reset Statistics	`wcagEnhancer.resetStatistics`	Reset statistics
AccessiMind: User Journey Scan	`wcagEnhancer.userJourneyScan`	Run user-journey accessibility analysis
AccessiMind: DOM Diff Risk	`wcagEnhancer.domDiffRisk`	Analyze DOM-change accessibility risk
AccessiMind: Design Token Guard	`wcagEnhancer.designTokenGuard`	Check design-token accessibility risk
AccessiMind: Component Memory	`wcagEnhancer.componentMemory`	Analyze component accessibility patterns
AccessiMind: Apply Last Fix To Similar	`wcagEnhancer.applyLastFixToSimilar`	Reuse the last fix pattern on similar code
AccessiMind: Generate Accessibility Test	`wcagEnhancer.generateA11yTest`	Generate an accessibility-focused test
AccessiMind: Generate PR Summary	`wcagEnhancer.generatePrSummary`	Generate a pull-request accessibility summary
AccessiMind: Regression Shield	`wcagEnhancer.regressionShield`	Run accessibility regression checks

Default keybindings:

Keybinding	Command
`Ctrl+Alt+W`	Analyze open file
`Ctrl+Alt+Shift+W`	Analyze selection

Configuration

Most settings are managed through the AccessiMind settings UI. The manifest-backed settings include:

Setting	Default	Notes
`wcagEnhancer.ai.provider`	`gemini`	`gemini`, `vscode-copilot`, `ollama`, or `codex-subscription`
`wcagEnhancer.ai.apiKey`	empty	Gemini API key
`wcagEnhancer.ai.ollamaUrl`	`http://localhost:11434`	Local Ollama endpoint
`wcagEnhancer.ai.ollamaModel`	`llama3`	Selected Ollama model
`wcagEnhancer.ai.codexPath`	`codex`	Codex CLI path
`wcagEnhancer.ai.codexTimeoutMs`	`180000`	Codex execution timeout
`wcagEnhancer.aiModels.selectedModel`	`gemini-2.5-flash`	Active model id
`wcagEnhancer.wcagLevel`	`AA`	Target WCAG level
`wcagEnhancer.language`	`en`	`auto`, `en`, or `tr`
`wcagEnhancer.strictMode`	`false`	Enables stricter analysis checks
`wcagEnhancer.contextAwareAnalysis`	`true`	Evaluates surrounding code context
`wcagEnhancer.customRulesPath`	empty	Optional markdown file with local rules
`wcagEnhancer.analysisDisabilityFocus`	`[]`	Optional disability-focus groups
`wcagEnhancer.performance.promptMaxChars`	`8000`	Prompt-size control
`wcagEnhancer.performance.maxScanSize`	`500000`	WCAG metadata scan limit
`wcagEnhancer.chatGptIntegration.enabled`	`false`	Enables ChatGPT handoff settings

Requirements

VS Code 1.93.0 or newer.
Node.js/npm for local development.
Gemini provider: Google Gemini API key.
VS Code Copilot provider: active GitHub Copilot access.
Ollama provider: local Ollama server.
Codex Subscription provider: Codex CLI installed and signed in with a ChatGPT/Codex account.
Browser bridge features: Chrome or Edge, with optional remote debugging for attach mode.

Development

Install dependencies:

npm install

Compile TypeScript:

npm run compile

Run unit tests:

npm run test:unit

Build the extension bundle:

npm run build

Package a VSIX:

npm run package

Main source areas:

Path	Purpose
`src/extension.ts`	Extension composition root
`src/extensionActions.ts`	Core command handlers and editor actions
`src/extensionCommands.ts`	Command registration
`src/extensionBootstrap.ts`	Settings/bootstrap helpers
`src/core/`	WCAG analysis and improvement logic
`src/infrastructure/providers/`	AI provider implementations
`src/agent/`	Agent session planning and apply workflow
`src/innovation/`	Advanced accessibility workflows
`src/views/`	VS Code webview and settings UI surfaces
`src/utils/`	Browser bridge, settings, runtime, and provider utilities
`browser-extension/accessimind-companion/`	Browser companion extension

Language Support

English UI and AI responses.
Turkish UI and AI responses.
auto language mode follows the VS Code environment where supported.

Repository

GitHub: sarperarikan/accessimind-vscode
Issues: github.com/sarperarikan/accessimind-vscode/issues
WCAG 2.2 quick reference: w3.org/WAI/WCAG22/quickref

License

MIT License. See LICENSE for details.

Author

Sarper Arikan
GitHub: @sarperarikan
Email: sarperarikan@gmail.com