diff --git a/.nav.yml b/.nav.yml index c611f654..6e9d576e 100644 --- a/.nav.yml +++ b/.nav.yml @@ -8,9 +8,8 @@ nav: - ':custom-terminal: CLI': code-reviews/cli - ':custom-database: External Knowledge': code-reviews/external-knowledge - ':custom-settings: Configuration': code-reviews/configuration - - 'Video Library': code-reviews/video-library.md + - ':custom-history: Changelog': code-reviews/changelogs.md - ':custom-faq: FAQ': code-reviews/faq.md - ':custom-alert: Troubleshooting': code-reviews/troubleshooting.md - - 'Reference': code-reviews/reference + - ':custom-file: Reference': code-reviews/reference.md - 'AI Resources': ai-resources.md - - 'Changelog': code-reviews/changelogs.md diff --git a/code-reviews/.nav.yml b/code-reviews/.nav.yml index a7a2d864..41d95a23 100644 --- a/code-reviews/.nav.yml +++ b/code-reviews/.nav.yml @@ -6,7 +6,6 @@ nav: - 'CLI': cli - 'External Knowledge': external-knowledge - 'Configuration': configuration - - 'Video Library': video-library.md - 'FAQ': faq.md - 'Troubleshooting': troubleshooting.md - - 'Reference': reference + - 'Reference': reference.md diff --git a/code-reviews/changelogs.md b/code-reviews/changelogs.md index 148a86b7..301fd0b7 100644 --- a/code-reviews/changelogs.md +++ b/code-reviews/changelogs.md @@ -2,7 +2,6 @@ title: kluster Code Reviews Changelog description: A chronological record of kluster.ai Code Reviews product updates, including new features, bug fixes, and improvements across all releases. categories: Changelog -footer_nav: true --- # Product changelog diff --git a/code-reviews/configuration/options.md b/code-reviews/configuration/options.md index a5211a90..0056f951 100644 --- a/code-reviews/configuration/options.md +++ b/code-reviews/configuration/options.md @@ -71,5 +71,5 @@ The default configuration uses **Deep** for human-driven workflows and **Instant ## Next steps - [Create custom rules](/code-reviews/configuration/rules/): Add project-specific development standards. -- [View MCP tools reference](/code-reviews/reference/mcp-tools/): Understand the technical API details. +- [View MCP tools reference](/code-reviews/reference/#mcp-tools): Understand the technical API details. - [Installation guide](/code-reviews/get-started/installation/): Set up Code Reviews in your IDE. diff --git a/code-reviews/ide-reviews/ai-generated-code/examples/cursor-firebase-nextjs.md b/code-reviews/ide-reviews/ai-generated-code/examples/cursor-firebase-nextjs.md index c41da9dd..899aa354 100644 --- a/code-reviews/ide-reviews/ai-generated-code/examples/cursor-firebase-nextjs.md +++ b/code-reviews/ide-reviews/ai-generated-code/examples/cursor-firebase-nextjs.md @@ -258,4 +258,4 @@ Even with clear prompts and detailed plans, AI execution can go wrong. Code Revi The more complex the task, the more valuable this real-time verification becomes. -**Learn more**: Explore our [MCP tools reference](/code-reviews/reference/mcp-tools/) to understand all issue types and priority levels that Code Reviews monitors. +**Learn more**: Explore our [MCP tools reference](/code-reviews/reference/#mcp-tools) to understand all issue types and priority levels that Code Reviews monitors. diff --git a/code-reviews/ide-reviews/ai-generated-code/examples/vscode-admin-endpoint.md b/code-reviews/ide-reviews/ai-generated-code/examples/vscode-admin-endpoint.md index c1d3ff10..5f4ad9b9 100644 --- a/code-reviews/ide-reviews/ai-generated-code/examples/vscode-admin-endpoint.md +++ b/code-reviews/ide-reviews/ai-generated-code/examples/vscode-admin-endpoint.md @@ -140,4 +140,4 @@ Admin endpoints require special security attention that AI often misses: [Code Reviews](/code-reviews/review-modes/) acts as your security safety net, catching vulnerabilities that look functional but hide critical flaws. The more powerful the operation, the more critical this protection becomes. -**Learn more**: Explore our [MCP tools reference](/code-reviews/reference/mcp-tools/) to understand all vulnerability types that Code Reviews monitors. +**Learn more**: Explore our [MCP tools reference](/code-reviews/reference/#mcp-tools) to understand all vulnerability types that Code Reviews monitors. diff --git a/code-reviews/ide-reviews/human-written-code/on-demand-reviews.md b/code-reviews/ide-reviews/human-written-code/on-demand-reviews.md index a3f85da9..d86a0e60 100644 --- a/code-reviews/ide-reviews/human-written-code/on-demand-reviews.md +++ b/code-reviews/ide-reviews/human-written-code/on-demand-reviews.md @@ -133,5 +133,5 @@ You can customize how on-demand reviews work in your [configuration options](/co ## Next steps -- **[MCP Tools Reference](/code-reviews/reference/mcp-tools/)**: Deep dive into all MCP tools and parameters. +- **[MCP Tools Reference](/code-reviews/reference/#mcp-tools)**: Deep dive into all MCP tools and parameters. - **[Configuration Options](/code-reviews/configuration/options/)**: Customize Code Reviews behavior for your workflow. diff --git a/code-reviews/reference.md b/code-reviews/reference.md new file mode 100644 index 00000000..53920a0d --- /dev/null +++ b/code-reviews/reference.md @@ -0,0 +1,162 @@ +--- +title: MCP Tools Reference +description: Learn how kluster.ai Code MCP tools work, including parameters, response formats, issue categories, and settings for real-time code reviews. +categories: Reference +--- + +# MCP Tools Reference + +This page provides the technical reference for the [kluster.ai](https://www.kluster.ai/){target=\_blank} Code MCP server, including available tools, request parameters, and the response format returned by code reviews. Use it as a quick lookup when working in Cursor, Claude Code, or any MCP-compatible client. + +## Available tools + +The Code MCP server provides review tools for checking code quality and security. These tools enable real-time code reviews directly within your IDE through MCP integration. + +- **`kluster_code_review_auto`**: Automatically reviews code quality and detects bugs, including logic errors, security issues, and performance problems. Triggers automatically when code is generated or modified. Best for real-time reviews during active coding sessions, analyzing changes in context of the full conversation and related files. +- **`kluster_dependency_check`**: Validates the security and compliance of packages and dependencies. Triggers automatically before package installations or when package files are updated. Best for preventing vulnerable or non-compliant third-party libraries from entering your codebase before installation. +- **`kluster_code_review_manual`**: Manually reviews specific files when explicitly requested by the user (e.g., "review this file", "check for bugs"). Best for auditing existing code, reviewing specific modules, or getting fix recommendations for individual files. + +This page documents the parameters and response formats you'll see when using these tools in Cursor, Claude Code, or any MCP-compatible client. + +## Parameters + +=== "Automatic Reviews and Dependency Check" + + These tools analyze code changes and dependencies to detect bugs, security vulnerabilities, and other quality issues. Used for AI-generated code. + + ???+ interface "Parameters" + + `code_diff` ++"string"++ ++"required"++ + + Unified diff format showing the actual changes (additions and subtractions) made to files. Use standard diff format with `--- filename` and `+++ filename` headers, followed by `@@ line numbers @@`, and `+` for additions, `-` for deletions. In MCP environments, this is often auto-extracted from IDE history. + + --- + + `user_requests` ++"string"++ ++"required"++ + + A chronological sequence of all user messages and requests in this conversation thread, with the current request (that triggered this assistant turn) clearly marked. Format: Previous requests as numbered list, then current request marked with `>>> CURRENT REQUEST: [request text]`. In MCP environments, this is often auto-extracted from conversation history. + + --- + + `modified_files_path` ++"string"++ ++"required"++ + + Full absolute paths of modified files separated by `;`. + + --- + + `chat_id` ++"string"++ ++"optional"++ + + Session identifier returned by previous tool calls. Used to maintain context across multiple review requests. + +=== "On-demand Review Tool" + + The on-demand review tool is triggered when explicitly requested. Used for both AI-generated code (when you ask your AI to review code) and human-written code (when you use the extension UI). + + ???+ interface "Parameters" + + `user_requests` ++"string"++ ++"required"++ + + Chronological sequence of user messages with current request marked as `>>> CURRENT REQUEST:`. Unlike automatic reviews, this parameter is NOT auto-extracted in MCP environments and must be explicitly provided. + + --- + + `modified_file_path` ++"string"++ ++"required"++ + + Full absolute path of the single file to review. This tool can only check one file per call. + + --- + + `need_fixes` ++"boolean"++ ++"required"++ + + Set to `true` if user requested fixes, `false` if only requesting issue detection. + + --- + + `chat_id` ++"string"++ ++"optional"++ + + Session identifier returned by previous tool calls. Used to maintain context across multiple review requests. + +## Response format + +All kluster.ai code review tools return the same response structure. This section documents the fields you'll see in review responses. + +### Response fields + +- **`isCodeCorrect`**: Boolean indicating if the code has issues. +- **`explanation`**: Summary of all issues found. +- **`issues`**: Array of detected problems with: + - **`type`**: Issue category (intent, semantic, knowledge, performance, quality, logical, security). + - **`severity`**: Impact level (critical, high, medium, low). + - **`priority`**: Execution priority (P0-P5). + - **`description`**: Brief issue summary. + - **`explanation`**: Detailed issue explanation. + - **`actions`**: Recommended fixes. +- **`priority_instructions`**: Execution rules for addressing issues. +- **`agent_todo_list`**: Prioritized list of fixes to apply. +- **`chat_id`**: Session identifier for maintaining context across requests. + +### Example response + +```json +{ + "isCodeCorrect": false, + "explanation": "Found 1 issue. 1 critical issue needs immediate attention.\n\nTODO:\n1. [CRITICAL] The implementation introduces a critical SQL injection vulnerability.", + "issues": [ + { + "type": "intent", + "severity": "critical", + "priority": "P0", + "description": "The implementation introduces a critical SQL injection vulnerability, which is an unacceptable security risk.", + "explanation": "The code constructs an SQL query using string concatenation with user input, which is the classic pattern for SQL injection. A function designed for database interaction should use parameterized queries.", + "actions": "Use parameterized queries or prepared statements to safely handle user input. For example: db.query('SELECT * FROM users WHERE id = ?', [userId])" + } + ], + "priority_instructions": "**PRIORITY EXECUTION RULES:**\n1. **INTENT Critical/High (P0-P1) get special priority**\n2. **All other issues sorted by severity** - Critical (P2) > High (P3) > Medium (P4) > Low (P5)\n3. **Never let lower priority issues override higher priority changes**", + "agent_todo_list": [ + "**EXECUTE IN THIS EXACT ORDER:**", + "", + "**Priority P0 - INTENT CRITICAL (HIGHEST PRIORITY):**", + "P0.1: The implementation introduces a critical SQL injection vulnerability - Use parameterized queries or prepared statements." + ], + "chat_id": "i8ct930591" +} +``` + +### Issue types + +| Type | Description | Example | +|------|-------------|---------| +| `intent` | Code doesn't match user request | Asked for sorting, got filtering | +| `semantic` | Meaning and type errors | Wrong variable type used | +| `knowledge` | Best practice violations | Not following conventions | +| `performance` | Performance issues | Inefficient algorithms | +| `quality` | Code quality problems | Poor naming, complexity | +| `logical` | Control flow errors | Off-by-one errors | +| `security` | Security vulnerabilities | SQL injection risks | + +### Severity levels + +| Level | Meaning | Action | +|-------|---------|--------| +| `critical` | Security vulnerability or breaking issue | Fix immediately | +| `high` | Significant bug or security concern | Fix before production | +| `medium` | Quality issue or minor bug | Should fix | +| `low` | Style or minor improvement | Optional fix | + +### Priority system + +Code review assigns priority levels to detected issues, helping you focus on the most critical problems first. The system automatically prioritizes based on issue type and severity. + +| Priority | Meaning | +|----------|---------| +| **P0-P1** | Intent issues (highest priority) - code doesn't match request | +| **P2** | Critical severity - must fix immediately | +| **P3** | High severity - should fix soon | +| **P4** | Medium severity - nice to fix | +| **P5** | Low severity - optional improvements | + +## Next steps + +- **[Configure settings](/code-reviews/configuration/options/)**: Customize review behavior for your needs. +- **[Installation](/code-reviews/get-started/installation/)**: Set up kluster.ai in your IDE. +- **[AI-generated code](/code-reviews/ide-reviews/ai-generated-code/automatic-reviews/)**: Learn how reviews work with AI assistants. diff --git a/code-reviews/reference/.nav.yml b/code-reviews/reference/.nav.yml deleted file mode 100644 index dea90e33..00000000 --- a/code-reviews/reference/.nav.yml +++ /dev/null @@ -1,4 +0,0 @@ -footer_nav: true -nav: - - 'MCP tools': mcp-tools.md - - 'Response schema': response-schema.md diff --git a/code-reviews/reference/mcp-tools.md b/code-reviews/reference/mcp-tools.md deleted file mode 100644 index 5fc418e4..00000000 --- a/code-reviews/reference/mcp-tools.md +++ /dev/null @@ -1,85 +0,0 @@ ---- -title: MCP Tools Reference -description: Learn how kluster.ai Code MCP tools work, including parameters, response formats, issue categories, and settings for real-time code reviews. -categories: Reference ---- - -# MCP Tools Reference - -The [kluster.ai](https://www.kluster.ai/){target=\_blank} Code MCP server provides review tools for checking code quality and security. These tools enable real-time code reviews directly within your IDE through MCP integration. - -## Available tools - -- **`kluster_code_review_auto`**: Automatically reviews code quality and detects bugs, including logic errors, security issues, and performance problems. Triggers automatically when code is generated or modified. Best for real-time reviews during active coding sessions, analyzing changes in context of the full conversation and related files. -- **`kluster_dependency_check`**: Validates the security and compliance of packages and dependencies. Triggers automatically before package installations or when package files are updated. Best for preventing vulnerable or non-compliant third-party libraries from entering your codebase before installation. -- **`kluster_code_review_manual`**: Manually reviews specific files when explicitly requested by the user (e.g., "review this file", "check for bugs"). Best for auditing existing code, reviewing specific modules, or getting fix recommendations for individual files. - -This page documents the parameters and response formats you'll see when using these tools in Cursor, Claude Code, or any MCP-compatible client. - -## Parameters - -=== "Automatic Reviews and Dependency Check" - - These tools analyze code changes and dependencies to detect bugs, security vulnerabilities, and other quality issues. Used for AI-generated code. - - ???+ interface "Parameters" - - `code_diff` ++"string"++ ++"required"++ - - Unified diff format showing the actual changes (additions and subtractions) made to files. Use standard diff format with `--- filename` and `+++ filename` headers, followed by `@@ line numbers @@`, and `+` for additions, `-` for deletions. In MCP environments, this is often auto-extracted from IDE history. - - --- - - `user_requests` ++"string"++ ++"required"++ - - A chronological sequence of all user messages and requests in this conversation thread, with the current request (that triggered this assistant turn) clearly marked. Format: Previous requests as numbered list, then current request marked with `>>> CURRENT REQUEST: [request text]`. In MCP environments, this is often auto-extracted from conversation history. - - --- - - `modified_files_path` ++"string"++ ++"required"++ - - Full absolute paths of modified files separated by `;`. - - --- - - `chat_id` ++"string"++ ++"optional"++ - - Session identifier returned by previous tool calls. Used to maintain context across multiple review requests. - -=== "On-demand Review Tool" - - The on-demand review tool is triggered when explicitly requested. Used for both AI-generated code (when you ask your AI to review code) and human-written code (when you use the extension UI). - - ???+ interface "Parameters" - - `user_requests` ++"string"++ ++"required"++ - - Chronological sequence of user messages with current request marked as `>>> CURRENT REQUEST:`. Unlike automatic reviews, this parameter is NOT auto-extracted in MCP environments and must be explicitly provided. - - --- - - `modified_file_path` ++"string"++ ++"required"++ - - Full absolute path of the single file to review. This tool can only check one file per call. - - --- - - `need_fixes` ++"boolean"++ ++"required"++ - - Set to `true` if user requested fixes, `false` if only requesting issue detection. - - --- - - `chat_id` ++"string"++ ++"optional"++ - - Session identifier returned by previous tool calls. Used to maintain context across multiple review requests. - -## Response format - -All code review tools return the same response structure. See the [Response Schema](/code-reviews/reference/response-schema/) for complete details. - -## Next steps - -- **[Response Schema](/code-reviews/reference/response-schema/)**: Understand the response format in detail -- **[Configure settings](/code-reviews/configuration/options/)**: Customize review behavior for your needs -- **[Installation](/code-reviews/get-started/installation/)**: Set up kluster.ai in your IDE diff --git a/code-reviews/reference/response-schema.md b/code-reviews/reference/response-schema.md deleted file mode 100644 index 5cc0a776..00000000 --- a/code-reviews/reference/response-schema.md +++ /dev/null @@ -1,90 +0,0 @@ ---- -title: Response Schema -description: Understand the response format from kluster.ai Code Reviews—issue structure, severity levels, priority system, and suggested fixes. -categories: Reference ---- - -# Response Schema - -All kluster.ai code review tools return the same response structure. This page documents the fields you'll see in review responses. - -## Response fields - -- **`isCodeCorrect`**: Boolean indicating if the code has issues. -- **`explanation`**: Summary of all issues found. -- **`issues`**: Array of detected problems with: - - **`type`**: Issue category (intent, semantic, knowledge, performance, quality, logical, security). - - **`severity`**: Impact level (critical, high, medium, low). - - **`priority`**: Execution priority (P0-P5). - - **`description`**: Brief issue summary. - - **`explanation`**: Detailed issue explanation. - - **`actions`**: Recommended fixes. -- **`priority_instructions`**: Execution rules for addressing issues. -- **`agent_todo_list`**: Prioritized list of fixes to apply. -- **`chat_id`**: Session identifier for maintaining context across requests. - -## Example response - -```json -{ - "isCodeCorrect": false, - "explanation": "Found 1 issue. 1 critical issue needs immediate attention.\n\nTODO:\n1. [CRITICAL] The implementation introduces a critical SQL injection vulnerability.", - "issues": [ - { - "type": "intent", - "severity": "critical", - "priority": "P0", - "description": "The implementation introduces a critical SQL injection vulnerability, which is an unacceptable security risk.", - "explanation": "The code constructs an SQL query using string concatenation with user input, which is the classic pattern for SQL injection. A function designed for database interaction should use parameterized queries.", - "actions": "Use parameterized queries or prepared statements to safely handle user input. For example: db.query('SELECT * FROM users WHERE id = ?', [userId])" - } - ], - "priority_instructions": "**PRIORITY EXECUTION RULES:**\n1. **INTENT Critical/High (P0-P1) get special priority**\n2. **All other issues sorted by severity** - Critical (P2) > High (P3) > Medium (P4) > Low (P5)\n3. **Never let lower priority issues override higher priority changes**", - "agent_todo_list": [ - "**EXECUTE IN THIS EXACT ORDER:**", - "", - "**Priority P0 - INTENT CRITICAL (HIGHEST PRIORITY):**", - "P0.1: The implementation introduces a critical SQL injection vulnerability - Use parameterized queries or prepared statements." - ], - "chat_id": "i8ct930591" -} -``` - -## Issue types - -| Type | Description | Example | -|------|-------------|---------| -| `intent` | Code doesn't match user request | Asked for sorting, got filtering | -| `semantic` | Meaning and type errors | Wrong variable type used | -| `knowledge` | Best practice violations | Not following conventions | -| `performance` | Performance issues | Inefficient algorithms | -| `quality` | Code quality problems | Poor naming, complexity | -| `logical` | Control flow errors | Off-by-one errors | -| `security` | Security vulnerabilities | SQL injection risks | - -## Severity levels - -| Level | Meaning | Action | -|-------|---------|--------| -| `critical` | Security vulnerability or breaking issue | Fix immediately | -| `high` | Significant bug or security concern | Fix before production | -| `medium` | Quality issue or minor bug | Should fix | -| `low` | Style or minor improvement | Optional fix | - -## Priority system - -Code review assigns priority levels to detected issues, helping you focus on the most critical problems first. The system automatically prioritizes based on issue type and severity. - -| Priority | Meaning | -|----------|---------| -| **P0-P1** | Intent issues (highest priority) - code doesn't match request | -| **P2** | Critical severity - must fix immediately | -| **P3** | High severity - should fix soon | -| **P4** | Medium severity - nice to fix | -| **P5** | Low severity - optional improvements | - -## Next steps - -- **[MCP Tools Reference](/code-reviews/reference/mcp-tools/)**: Learn about tool parameters -- **[Configure settings](/code-reviews/configuration/options/)**: Customize what issues get flagged -- **[AI-generated code](/code-reviews/ide-reviews/ai-generated-code/automatic-reviews/)**: Learn how reviews work with AI assistants