Skip to main content

​
0.3.4 (Feb 3, 2026)

Neo’s biggest release yet: VPN connectivity, a new benchmark evaluation framework, and a runtime migration deliver enterprise-grade security testing with significantly improved performance across the board. New
  • Added Opus 4.6 and latest model updates for more capable AI reasoning
  • Introduced custom working memory so Neo retains richer context across tasks
  • Launched async task trigger API for non-blocking, parallel task execution
  • Added per-message stream status for granular real-time progress tracking
  • Added Neo credits validation and usage tracking system
  • Introduced secure credential management via the add_secret tool
  • Launched Benchmark Evaluation Framework with budget management and CLI tooling for measuring Neo’s performance
  • Added knowledge search and indexing tools to the sandbox agent
  • Introduced VPN connectivity with metrics tracking and export for testing internal networks
  • Added a high-performance Go-based indexing sidecar service
  • Introduced viewer role for read-only access to shared workspaces
  • Added Docker Compose support in sandbox environments
Improvements
  • Enhanced memory recall with direct message query support
  • Implemented automatic tool call recovery for more resilient agent execution
  • Migrated agent runtime to Bun for faster startup and execution
  • Improved NeoAgent with automatic model retry on failures
  • Fixed browser agent tools and session management
  • Fixed Python script execution and web image analysis tools
  • Fixed sub-agent creation and editing workflows
  • Fixed edge cases in token tracking with invalid numeric values
  • Added folder delete endpoint and cleaned up stale sandbox entries

​
0.3.3 (Jan 17, 2026)

Teams and third-party integrations arrive in Neo. Collaborate with your security team and connect directly to GitHub, Jira, Slack, and Linear from within your workflows. New
  • Launched Teams with multi-user collaboration support
  • Added Connectors for third-party integrations: GitHub, Jira, Slack, and Linear
  • Added database migration automation for smoother upgrades
  • Added infrastructure monitoring alerts for proactive issue detection
Improvements
  • Upgraded prompt system to YAML-based configuration with automatic fallback
  • Fixed tool caching and improved cache reliability
  • Fixed stream error handling for more stable long-running tasks
  • Fixed handling of auto-joined team members in filtering and invites
  • Added proper aborted status reporting for cancelled tasks
  • Fixed sub-agent and infrastructure timeout handling

​
0.3.2 (Jan 15, 2026)

Stability patch improving stream reliability for long-running security scans and complex multi-step operations. Improvements
  • Fixed stream chunk handling for improved reliability during extended scans

​
0.3.1 (Jan 14, 2026)

Enhanced admin capabilities with usage analytics, improved CI/CD pipeline, and browser proxy configuration for more flexible security testing scenarios. New
  • Added admin usage analytics dashboard API
  • Added tag-based proxy configuration for browser sessions
  • Added automated build testing, linting, and test coverage to the CI pipeline
  • Added data redaction tooling for privacy compliance
Improvements
  • Improved scheduled task context management
  • Split sandbox builds into production and development tracks for faster iteration
  • Fixed authentication key validation
  • Optimized semantic memory performance by disabling unnecessary global recalls
  • Added detailed logging for stream failure diagnostics

​
0.3.0 (Jan 12, 2026)

Security-first release introducing tool-level secret redaction and native ProjectDiscovery Cloud Platform integration for domain intelligence and reconnaissance. New
  • Introduced tool-level secret redaction: credentials are now automatically masked in task outputs
  • Added native PDCP Associated Domains integration for richer domain intelligence
Improvements
  • Cleaned up command output formatting in sandbox for more readable logs
  • Fixed stale connection issues in agent APIs

​
0.2.9 (Jan 11, 2026)

User organization improvements with a tagging system for better task and team management. New
  • Added user tagging system for organizing and filtering tasks and team members
Improvements
  • Improved message length handling for more accurate token counting

​
0.2.8 (Jan 10, 2026)

Infrastructure simplification removing external cache dependencies from sandbox management, improving system reliability and reducing operational complexity. New
  • Added automatic cleanup for stale sandbox records when backing services are unavailable
  • Added Firefox compatibility for cross-browser access
Improvements
  • Simplified sandbox architecture by removing external cache dependency
  • Fixed stream status handling for aborted tasks
  • Fixed message truncation with comprehensive test coverage
  • Fixed file execution commands

​
0.2.7 (Jan 8, 2026)

Admin API expansion enabling programmatic user management and improved audit trails with timestamps. New
  • Added user creation API endpoint for programmatic account management
  • Added created and updated timestamps to user records for audit trails
Improvements
  • Fixed sandbox retrieval operations

​
0.2.6 (Jan 6, 2026)

Feature flags infrastructure arrives, enabling controlled rollout of new Neo capabilities to specific users and teams. New
  • Added feature flag system for controlled capability rollout per user
Improvements
  • Fixed handling of corrupted temporary workspace files
  • Fixed sandbox health detection and automatic recovery

​
0.2.5 (Jan 5, 2026)

Rate limiting and authentication hardening for production deployments. New
  • Added token-based rate limiting middleware
  • Enabled admin authentication for internal service calls in production
Improvements
  • Cleaned up stream handling and enforced cache requirements for API services

​
0.2.4 (Jan 5, 2026)

Comprehensive error handling improvements across the scheduler and streaming infrastructure, making Neo more resilient during complex, long-running security operations. New
  • Added task lineage tracking via source task IDs
  • Added default model fallback for schedule creation
Improvements
  • Fixed model configuration naming for consistent AI provider communication
  • Updated internal SDK types for latest framework compatibility
  • Added request context to error logs for better debugging
  • Improved error logging with severity levels and richer context
  • Fixed stream controller race condition and improved error recovery
  • Fixed title generation failures
  • Prevented embedding token limit errors
  • Improved scheduler reliability and cleanup
  • Fixed database type inference in stream status updates

​
0.2.3 (Jan 4, 2026)

Task privacy controls migrate to the high-performance API backend, enabling users to control visibility and redact sensitive information from shared tasks. Improvements
  • Migrated task visibility and redaction features to the Go API for better performance
  • Added context to AI provider errors for easier troubleshooting
  • Fixed schedule lock and timeout issues

​
0.2.2 (Jan 2, 2026)

Native API key authentication launches, allowing programmatic access to Neo without browser-based sign-in flows. New
  • Introduced Neo-native API key authentication system
  • Added comprehensive API endpoint test coverage
Improvements
  • Fixed task details stream URLs

​
0.2.1 (Dec 31, 2025)

REST API cleanup completing the v2 endpoint migration with standardized patterns across all endpoints. Improvements
  • Completed REST API cleanup and standardization across all endpoints
  • Fixed prompt configuration sync in CI workflows
  • Fixed sub-agent stream consumption to properly resolve multi-step tasks

​
0.2.0 (Dec 27, 2025)

Major architecture milestone: user-facing APIs migrate from Node.js to Go, delivering significant performance improvements and setting the foundation for Neo’s scalability. New
  • Migrated user-facing APIs to Go for dramatically improved performance and lower latency
  • Standardized all REST endpoints to v2
  • Added chat rename endpoint
  • Added fine-grained tool streaming for Anthropic models
  • Unified development environment with single command startup
Improvements
  • Added Go API service containerization
  • Fixed shared chat visibility settings
  • Removed unused stream storage layers
  • Fixed stream handling for concurrent readers

​
0.1.0 (Dec 22, 2025)

Share your security research with the world. Public task sharing enables collaboration and knowledge sharing across the security community. New
  • Added public task sharing for collaborative security research
  • Upgraded agent framework packages for improved capabilities

​
0.0.9 (Dec 19, 2025)

Zero-latency resumable streams transform the user experience: reconnect to long-running security scans instantly without losing any progress. New
  • Launched zero-latency resumable streams for instant task resumption
  • Added real-time streaming of live task chunks
Improvements
  • Added background health checks with cooldown tracking for cache services
  • Fixed health checks to avoid blocking the event loop
  • Optimized cloud storage syncs with debouncing for better performance
  • Fixed abort flag handling
  • Fixed stream cleanup on disconnection
  • Improved chat loading speed by removing redundant reconstruction
  • Pinned agent framework version for stability

​
0.0.8 (Dec 16, 2025)

Stream infrastructure optimizations reducing cache overhead and improving real-time response delivery during security scans. Improvements
  • Improved stream chunk storage efficiency
  • Reduced cache polling from per-chunk to 1-minute intervals
  • Fixed stream recovery on reconnection

​
0.0.7 (Dec 16, 2025)

Observability controls enabling operators to fine-tune logging and monitoring behavior via environment configuration. New
  • Added environment variable controls for observability and tracing
Improvements
  • Fixed tool execution ordering
  • Fixed development build issues

​
0.0.6 (Dec 16, 2025)

Performance release dramatically improving chat and task loading times for a snappier, more responsive experience. Improvements
  • Significantly improved chat and task loading performance
  • Reorganized message storage for better scalability
  • Fixed output accumulation to prevent blocking during cloud storage syncs
  • Bundled fallback prompts in production image for reliability
  • Removed unused context compaction feature

​
0.0.5 (Dec 15, 2025)

Production container improvements and refined task usage tracking with per-message granularity for accurate billing and reporting. Improvements
  • Fixed production container bundling of fallback prompts
  • Refactored task usage to per-message tracking with improved API

​
0.0.4 (Dec 14, 2025)

Scheduled tasks launch: automate recurring security scans and workflows to run on your schedule without manual intervention. New
  • Launched scheduled tasks for automated recurring security operations
  • Added character-based token counting for accurate usage tracking
Improvements
  • Various agent framework maintenance improvements

​
0.0.3 (Dec 14, 2025)

Task usage tracking enables visibility into resource consumption, helping teams understand and optimize their security testing workflows. New
  • Added task usage tracking and API for monitoring resource consumption
Improvements
  • Fixed memory instance consistency across sessions

​
0.0.2 (Dec 12, 2025)

Prompt management integration brings experiment mode and improved prompt engineering capabilities to Neo’s agent system. New
  • Integrated prompt management system with experiment mode for A/B testing prompts
  • Added status field to chat API responses
  • Added configuration comparison workflow
Improvements
  • Migrated to modern linting for improved code quality
  • Fixed sub-agent manager issues
  • Fixed custom file type handling
  • Prevented email exposure in agent directory (only username prefix is returned)
  • Fixed file tracking and local file indexing
  • Improved file listing performance
  • Fixed stale file index detection
  • Moved file uploads to dedicated workspace directory
  • Increased chat endpoint rate limits for development

​
0.0.1 (Dec 4, 2025)

Neo launches as the agentic backend for ProjectDiscovery’s AI-powered security platform. Built with mandatory Nuclei template verification, this release establishes the foundation for autonomous cybersecurity operations. New
  • Launched Neo API, the agentic security testing backend
  • Added mandatory Nuclei template verification for all discovered vulnerabilities
  • Implemented resumable streams for uninterrupted long-running tasks
  • Added structured logging for full observability
  • Added curl, nmap, and comprehensive security toolchain to sandbox
  • Added GitHub code search integration
Improvements
  • Enhanced cybersecurity agent with optimized context and tool selection
  • Simplified container image by removing unnecessary build stages
  • Added sandbox bin directory to PATH for seamless tool access
  • Added comprehensive test infrastructure with mocking support
  • Fixed libpcap compatibility for network scanning tools

​
Pre-Beta Releases

The following releases document Neo’s early development phase, originally codenamed “Loki API”. These pre-beta releases laid the groundwork for the production-ready platform.

​
pre-beta.6 (Nov 26 - Dec 3, 2025)

Production hardening phase with isolated cloud infrastructure, terminal customization, and graceful shutdown handling, preparing for the first stable release. New
  • Added per-user isolated sandbox environments on dedicated cloud infrastructure
  • Added custom display name and hostname in terminal sessions
  • Added Anthropic reasoning support with thinking mode
  • Implemented bidirectional cloud storage syncs for persistent file access
  • Added workspace auto-organization system for sandbox files
  • Added Docker and additional security tools to sandbox environments
Improvements
  • Implemented graceful shutdown handling to prevent data loss during cleanup
  • Revised recon workflow with validation and ASN filtering
  • Added feature access control caching for faster permission checks
  • Fixed cloud volumes and region configuration
  • Increased sandbox agent step limits for complex multi-step operations

​
pre-beta.5 (Nov 15 - Nov 25, 2025)

Knowledge base and vector search capabilities arrive, enabling semantic search across security research, vulnerability data, and uploaded documents. New
  • Integrated vector database for semantic search across knowledge bases
  • Added initial knowledge base APIs and tools
  • Added real-time terminal access via Neo Terminal Socket
  • Added Opus 4.5, Haiku 4.5, and updated Grok 4.1 models
  • Added knowledge indexing service for automated content processing
  • Added active session metrics for infrastructure monitoring
Improvements
  • Improved JSON/JSONL document chunking with integration tests
  • Fixed shared chat message functionality
  • Cleaned up unused database tables and optimized queries

​
pre-beta.4 (Nov 1 - Nov 14, 2025)

Agent directory and multi-model expansion with GPT 5.1 series, memory recall improvements, and native ProjectDiscovery tool integration. New
  • Launched Agent Directory with sharing and one-click installation
  • Added GPT 5.1 model series support
  • Added native ProjectDiscovery tools with auto-injected API credentials
  • Added chat mode with Haiku for fast, cost-efficient conversations
  • Added Grok 4 Fast and Kimi K2 models
  • Added analyze_web_images tool for URL-based image analysis
  • Added task snapshot redaction for privacy-sensitive sharing
  • Added direct agent execution and multi-agent filtered orchestration
Improvements
  • Added memory recall with token limits for better context management
  • Added Neo identity and time awareness in system context
  • Fixed token counter memory leak
  • Fixed sub-agent inline script execution

​
pre-beta.3 (Oct 15 - Oct 31, 2025)

Browser automation matures with cloud browser integration, persistent sessions, and real-time log capture for comprehensive web application security testing. New
  • Added cloud browser automation for web application testing
  • Added persistent browser sessions per user
  • Added real-time browser log capture with scope filtering
  • Added browser_extract_endpoints tool with session and page modes
  • Added per-user static IP proxy support for browser sessions
  • Added multipart upload support for large files (up to 1GB)
  • Added preview environment support
Improvements
  • Configured browser automation to use Claude Sonnet 4.5
  • Added file management endpoints with improved API consistency
  • Made chat sharing functional with unlisted visibility for custom agents
  • Fixed browser tool errors and improved error handling
  • Fixed cost calculation accuracy

​
pre-beta.2 (Sep 15 - Oct 14, 2025)

Agent framework integration brings structured orchestration, multi-model provider support, and distributed tracing for full observability into Neo’s AI operations. New
  • Integrated structured agent orchestration framework
  • Added multi-model provider support for flexible AI model selection
  • Added distributed tracing with observability exporter
  • Added Gemini 3 Pro support
  • Added API-based authentication with API versioning
  • Added XSS sub-agent and xss_contexts tool for vulnerability testing
  • Added sqlmap integration for SQL injection testing
  • Added web research tools
  • Added semantic search capabilities
  • Implemented database-backed resumable streams for reliability
  • Added global tool timeout wrapper
Improvements
  • Fixed undefined text issues in AI SDK
  • Fixed websocket retry logic
  • Improved logging severity levels and error context

​
pre-beta.1 (Aug 15 - Sep 14, 2025)

Agent swarm architecture takes shape with dynamic sub-agents, working memory, and the foundation for multi-agent security operations. New
  • Introduced agent swarm pattern for multi-agent orchestration
  • Added dynamic sub-agents with configurable concurrency limits
  • Implemented sub-agent working memory for context persistence
  • Added summary agent with memory access for synthesizing findings
  • Added error tracking integration for reliability monitoring
  • Added AI tracing for agent observability
  • Added Anthropic thinking mode support (experimental)
  • Added execute_script tool with ffuf integration for fuzzing
  • Added API security tools and improved code search
  • Reorganized out-of-band testing agents with separated CVE intelligence
Improvements
  • Fixed browser agent stability
  • Fixed MCP memory leak
  • Improved semantic memory size management

​
pre-beta.0 (Jul 29 - Aug 14, 2025)

The genesis of Neo. Initial cybersecurity agent with sandboxed execution, tool calling, and the first ProjectDiscovery integrations, laying the foundation for autonomous security testing. New
  • Initial project setup with core API structure
  • Added structured logging infrastructure
  • Implemented tool calling with heartbeat monitoring
  • Added sandboxed execution environment on cloud infrastructure
  • Added XSS and Open Redirect vulnerability verification
  • Added out-of-band testing (OAST) tools
  • Added XSS knowledge search capabilities
  • Added OpenAPI server support for extensibility
  • Added browser automation in sandbox
  • Added GPT-5 and Gemini 2.5 Pro model support
  • Added context compaction for continuous conversations
  • Added resumable streams foundation
  • Added mandatory Nuclei template verification
  • Added comprehensive Docker security toolchain
Improvements
  • Fixed libpcap compatibility for network scanning
  • Added curl and nmap to sandbox toolchain