Skip to content

CHANGELOG.md

Latest commit

 

History

History
180 lines (137 loc) · 6.76 KB

CHANGELOG.md

File metadata and controls

180 lines (137 loc) · 6.76 KB

Changelog

All notable changes to MCPLint will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.4.0] - 2025-12-14

Added

  • Advanced Prompt Engineering (src/ai/prompt_templates.rs)

    • VulnCategory enum for 8 vulnerability categories (Injection, Authentication, Cryptographic, etc.)
    • FewShotExample struct with category-specific examples
    • AdvancedPromptBuilder with chain-of-thought reasoning
    • Confidence scoring in AI explanations
    • Category-specific system prompts for better accuracy

  • Neo4j Knowledge Graph Integration (optional --features neo4j)

    • SecurityKnowledgeGraph for storing and querying vulnerability findings
    • Vector similarity search using cosine distance
    • VoyageEmbedder for code-optimized embeddings (voyage-code-2, 1536 dimensions)
    • CWE/CVE knowledge retrieval
    • Cross-server vulnerability pattern detection

  • Optional Feature Flags

    • neo4j - Neo4j graph database for vulnerability knowledge base
    • redis - Redis distributed cache backend

  • Async Performance Optimization

    • Parallel security detector execution in scanner engine using futures::stream::buffer_unordered
    • Parallel tool/resource/prompt validation using Rayon parallel iterators
    • Parallel JSON Schema validation for large tool sets
    • Design document at docs/async-optimization-plan.md

  • Streaming Scan Results (src/scanner/streaming.rs)

    • FindingStream for memory-efficient consumption of scan findings
    • FindingProducer for streaming findings from scanner to consumers
    • ScanEngine::scan_streaming() method for streaming scan API
    • Backpressure support via bounded tokio channels
    • Summary accumulation during streaming (no need to hold all findings)
    • ~99% memory reduction for large scans (10K findings: 50MB → 50KB)
    • Backward compatible: collect_all() for consumers that need all findings
    • 19 new unit tests for streaming functionality
    • Design document at docs/memory-optimization-plan.md

  • Enhanced Error Context

    • All file I/O operations now include contextual error messages using anyhow::Context
    • Error messages now specify which file/directory failed (e.g., "Failed to read config from /path/to/file")
    • Improved debugging experience with clear operation context in error chain
    • Applied to: init.rs, filesystem.rs, corpus.rs, fingerprint.rs, completions.rs, main.rs
    • Design document at docs/error-context-plan.md

Changed

  • AI providers now support advanced prompts with use_advanced_prompts flag
  • Ollama provider defaults to simplified prompts for better local model performance
  • Scanner run_advanced_security_checks now async with parallel detector execution
  • Validator run_protocol_rules and run_schema_rules use Rayon parallel iterators

Fixed

  • Clippy lints for CI compatibility (needless borrows, single char push, duplicated cfg attributes)
  • Fuzzer session test overflow on Windows
  • ExplainEngine integration test flakiness with retry logic
  • Additional clippy lints (field_reassign_with_default, useless_vec, default_constructed_unit_structs)

Tests

  • Added 112+ new tests across modules
  • Total test count: 3,293 passing tests (19 new streaming tests)
  • Neo4j integration tests (require live connection)

[0.3.1] - 2025-12-13

Added

  • HTML Output Format
    • Rich HTML reports with severity distribution charts
    • Finding cards with detailed information
    • Remediation guidance integration
    • Responsive design for all devices
    • Available via --format html on scan command
    • Integrated into interactive scan wizard

[0.3.0] - 2025-12-12

Added

  • Interactive Mode

    • Scan wizard with server selection, profile choice, and category filtering
    • Fuzz wizard with profile, duration, workers, and corpus configuration
    • Init wizard for guided configuration file creation
    • Explain wizard with AI provider, audience level, and severity filtering
    • Automatic TTY and CI environment detection
    • FuzzySelect for intuitive server selection

  • Init Command Enhancements

    • GitHub Actions workflow generation (.github/workflows/mcplint.yml)
    • Automatic .gitignore entry for .mcplint-cache/
    • Wizard-driven configuration with sensible defaults

Changed

  • Server argument now optional for scan, fuzz, and explain commands (wizard activates)
  • OutputFormat and ScanProfile types consolidated in cli module
  • Improved module organization for library consumers

Tests

  • Added interactive_tests.rs with 30 integration tests
  • Test coverage for wizard result structs and output formats
  • AI provider and audience level variant tests
  • CI environment detection tests

[0.2.0] - 2025-12-12

Added

  • Phase 1: Smart Context Detection

    • Automatic detection of TTY, CI, and plain output modes
    • NO_COLOR environment variable support
    • Unicode/ASCII fallback based on terminal capabilities

  • Phase 2: Progress Indicators

    • Real-time progress bars for scan operations
    • Connection spinners with phase tracking
    • Multi-server progress tracking

  • Phase 3: Output Formatting

    • Unified Printer API for consistent CLI output
    • Security-themed color system for severity levels
    • Structured output with separators, headers, key-value pairs

  • Phase 4: Enhanced Error Handling

    • Miette-based diagnostic errors with source context
    • "Did you mean?" suggestions using Jaro-Winkler similarity
    • Contextual help for common errors (connection, timeout, config)

  • Phase 5: Shell Completions

    • Dynamic shell completions for bash, zsh, fish, PowerShell
    • Server name completion from Claude Desktop config
    • Profile and format completion with descriptions

  • Phase 6: Watch Mode & CI Integration

    • Differential watch mode showing new/fixed issues
    • Debounced file watching with configurable interval
    • Enhanced SARIF output for GitHub Code Scanning

  • Multi-Server Scanning

    • Parallel scanning of multiple MCP servers
    • Configurable concurrency with semaphore control
    • Combined SARIF output for CI/CD pipelines
    • Aggregated statistics and severity counts

Changed

  • Improved CLI user experience across all commands
  • Better error messages with actionable suggestions
  • Optimized startup time (<20ms)

Performance

  • Startup time: ~18ms (target <100ms)
  • Binary size: 8.0MB
  • Test suite: 4,519 tests in ~2.2s

[0.1.0] - 2025-12-01

Added

  • Initial release
  • MCP protocol validation (56 rules)
  • Security scanning with 15+ detection rules
  • Coverage-guided fuzzing
  • AI-assisted vulnerability explanation
  • Multi-backend caching (memory, filesystem, Redis)
  • SARIF, JUnit, GitLab output formats
  • Watch mode for development
  • Doctor command for environment checks