Snapshot 253724
Normalized text
Scripts and page chrome removed; this is what change detection compares.
Blitzy Public Build: Bring An Idea, Get Up To 100k Lines Of Code Generation, At No Cost Read the blog Blitzy Public Build: Bring An Idea, Get Up To 100k Lines Of Code Generation, At No Cost Read the blog Blitzy Public Build: Bring An Idea, Get Up To 100k Lines Of Code Generation, At No Cost Read the blog SOC 2 type II compliant ISO 27001 certified Built for trust. Built for trust. Built for trust. Blitzy is designed for the enterprise, with security and compliance at the core from day one. Visit our trust center Visit our trust center No training on your code Blitzy never trains on user code. And we never store your code directly. Instead, solely store embeddings, ensuring your IP remains fully protected. Secure end-to-end encryption All customer data is encrypted in transit and at rest. We use industry-leading encryption protocols to ensure complete data security. Air-gapped code generation Our code-generation platform is completely isolated, not accessible via our user-facing interface or publicly over the internet, creating an air-gap for maximum security. Inbound-only VPC architecture Blitzy’s platform is an inbound-only, VPC-based system that never initiates outbound requests. The platform also has no publicly exposed endpoints, dramatically reducing the attack surface. Flexible deployment options Flexible deployment options Flexible deployment options Six ways to run Blitzy, from fully managed SaaS to air gapped premises. BLITZY · DEPLOYMENT ARCHITECTURE Where Blitzy runs: six deployment modes Every mode runs the same platform: a web app and API, a prompt sanitizer and agentic orchestration in the control plane, single-use job VMs in the execution plane, and a per-customer knowledge graph (a hybrid graph+vector database) in the data plane. What changes is who hosts each part, where your code is processed, and which network paths cross your perimeter. Your environment (solid border) Blitzy-hosted or external service (dashed) Inside your perimeter Hosted by Blitzy MODE 1 · CLOUD SaaS 2 · CLOUD STS 3 · HYBRID CLOUD Hybrid VPC 4 · CLOUD (YOURS) VPC Blackbox 5 · CLOUD (YOURS, GOV) VPC Blackbox Gov 6 · ON-PREMISES On-premises Blackbox Multi-tenant, Blitzy-managed Single-tenant, Blitzy-managed STS plus the Blitzy Client, a packet router in your network Full stack in your private cloud, customer-managed Full stack inside your FedRAMP High or DoD IL4/IL5 cloud boundary, customer-managed Physical appliance at your site Hosted and operated by Blitzy, on shared infrastructure with a public cloud provider Blitzy, in a cloud project and VPC dedicated to you Blitzy runs the STS project; you run the Blitzy Client router in your DMZ You, in your own cloud provider account You, in your government cloud account, inside your authorization boundary You, on Blitzy-supplied hardware inside your data center Control plane (web app, API, orchestration) Shared Blitzy platform Your dedicated project Your dedicated STS project Your Kubernetes cluster, installed by Helm Your Kubernetes cluster, installed by Helm from images in your own registry On the appliance Execution plane (clone, build, test) One single-use VM per job, never reused One single-use VM per job, dedicated to you Job VMs in STS reach your internal Git-based SCM, packages and test databases through the Blitzy Client Single-use job VMs in your account Hardened single-use job VMs in your account, destroyed after each task Single-use job VMs on the appliance Data plane (graph, specs, secrets) Per-company knowledge graph and bucket, isolated by tenant ID Dedicated knowledge graph, storage and secrets store; customer-managed keys optional Dedicated, in STS; credentials encrypted with your key (CMEK) Your cluster, storage, secrets vault and keys Your cluster, storage and vault; FIPS 140-validated encryption, keys in your KMS or HSM On the appliance AI model inference Frontier LLM providers; zero data retention, no training Same providers, from your dedicated egress IPs Same providers, from STS; Blitzy’s keys or, optionally, your own Model endpoints you approve, through your egress controls Government-cloud LLM endpoints you approve; all other model APIs disabled Open-weight models served on the appliance’s GPUs What crosses your perimeter Repository clone into Blitzy over HTTPS; branch and pull request back Same, to your dedicated project only One outbound link from the Blitzy Client (TCP 443, mTLS). Blitzy never dials in. Model calls and release pulls, through egress you control Nothing by default: no call-home, model traffic stays in your cloud; signed packages you scan and approve No outbound connection required; updates arrive as signed packages Ways to connect to your systems Public endpoints with IP allowlist, private link, or site-to-site VPN Same three options, terminating in your dedicated VPC Blitzy Client link over the internet or inside your IPsec VPN Private, inside your network Private, inside your network; SSO through your government IdP with MFA and CAC/PIV Private, inside your network Compliance posture Covered by Blitzy’s SOC 2 Type II report and ISO 27001 certification Covered by Blitzy’s SOC 2 Type II report and ISO 27001 certification, on infrastructure dedicated to you STS is covered by Blitzy’s SOC 2 Type II report and ISO 27001 certification; the Blitzy Client runs under your controls Runs under your cloud controls. Blitzy holds a SOC 2 Type II report and ISO 27001 certification. Runs in your FedRAMP High or DoD IL4/IL5 authorized cloud, under your agency’s ATO. Runs under your site controls. Blitzy holds a SOC 2 Type II report and ISO 27001 certification. Personnel access Blitzy support staff can reach the platform to help you; nothing reaches into your network Blitzy support staff, scoped to your dedicated project Blitzy staff work in STS only; nothing dials into your network Only the access you grant No standing access to your data plane; sessions you approve, time-limited, logged, US persons only No standing access; only sessions you allow Audit and supply chain Sign-in and activity events export to your SIEM; every change is a pull request Project-scoped audit logs, exportable to your SIEM Same as STS, plus a signed Blitzy Client package Logs and traces stay in your account; signed images and Helm chart Logs and agent traces stay in your boundary, export to your SIEM; signed, scanned images, Dockerfiles, Helm chart and SBOM Logs stay on the appliance; signed update packages Availability Standard On request Deal-specific Deal-specific Deal-specific Deal-specific In every mode: no AI agent has a network path into your environment, prompts are screened before any model sees them, agents receive task-scoped code snippets rather than whole repositories, and every change arrives as a pull request under your branch protection. Deployment architecture and data flow Deployment architecture and data flow SaaS STS Hybrid VPC VPC Blackbox VPC Blackbox Gov On-premises Blackbox SaaS STS Hybrid VPC VPC Blackbox VPC Blackbox Gov On-premises Blackbox SaaS STS Hybrid VPC VPC Blackbox VPC Blackbox Gov On-premises Blackbox Frequently asked questions Frequently asked questions What is Blitzy? Blitzy enables development teams to transform six-month software projects into six-day turnarounds using Blitzy OS, an agentic platform that enables thousands of AI Agents to 'think' and cooperate for hours to bulk build software with precision. The platform builds everything AI can deliver in a precise manner, around 80% of any roadmap or new product, supplemented with a human engineering guide to complete the remaining 20% needed for production. With over 27 patents and counting, Blitzy is actively hiring PhDs and senior developers in Cambridge, MA who have a passion for building AI that leverages 'System 2 Thinking' to solve problems at inference. Who is Blitzy for? Enterprises that aim to dramatically accelerate their software development velocity, development agencies with enterprise clients, development teams with complex existing products, and individuals looking to accelerate their own velocity on complex builds. How does Blitzy’s technology work? Our patent-pending code ingestion framework maps a curated selection of robust, reliable, and secure open source software libraries that we track by version and update frequently. Combined with our proprietary code generation technology that specializes on enforcing enterprise-class software policies, Blitzy far exceeds the utility of typical chatbots and co-pilots in creating enterprise-grade software at scale. Is Blitzy a coding co-pilot? Nope. Blitzy surpasses traditional co-pilots with its ability to autonomously generate nearly-complete code repositories, not just snippets. It features a daily-refreshed knowledge base, avoiding the pitfalls of outdated information. Blitzy's proprietary codebase representation system enables deep understanding of generated code, offering highly contextual and relevant suggestions for your entire repository. What's my role in Blitzy's development process? Your team is responsible for bringing the requirements, and as an approver during the technical specification stage. We ask you to edit/approve the Technical Specification. The document is editable, so you can edit and approve to get exactly what you had in mind. How does Blitzy decide which tasks to delegate to human developers? Blitzy's multi-agent system is meticulously and rigorously trained to know what it can accomplish, and what needs to be left for the human engineers. This ensures you only receive quality code and have a clear picture of remaining tasks. Does Blitzy do more than just autonomous code generation? Yes. Blitzy is a comprehensive platform that provides end-to-end development assistance. We support the entire development lifecycle by taking descriptive inputs and generating software requirements documents, technical design, code structure, and generative code within repos for your product. Is this high quality and secure? Quality and security matter deeply to us — and they were our biggest frustration with the copilots already on the market. That frustration is what led us to build something different: a system designed to meet enterprise standards from the start. Every piece of work passes through multiple QA agents that review each other's output before any code reaches you, so what you receive is held to a consistent quality bar rather than the variable output typical of single-pass code generation. We deliver enterprise-grade code repositories. As with any code entering your environment — written by humans or AI — your team should still run its own QA, QC, and security testing before deployment. We build to a high standard and give your reviewers a strong starting point; final validation stays with the team that owns the production environment. After submitting my prompt, Blitzy added functionality in my tech spec that I did not expect. What do I do? The system defaults to taking advantage of all technology upgrades when modernizing or upgrading to the latest technology stack. For example, if you specify an upgrade to Java 21, the system will by default implement virtual threads, as it's generally seen as a superior technical approach. If you do not want this, you must simply tell the system to 'make as few changes as possible to achieve the desired request'. Being as specific as possible about what functionality is (and is not) desired helps yield results that will align with expectations. What do Blitzy agents rely on as a source of truth to represent my existing codebase? Blitzy agents rely on the actual source code of your existing codebase—not the Tech Spec documentation—when performing refactors or extending functionality. However, an accurate Tech Spec significantly aids the system's efficiency in querying the underlying representation of the code. Therefore, investing time to ensure the Tech Spec reflects the core features of the application will yield expectation-aligned results and will save time with last-mile development. Can Blitzy work with existing products and code bases? Yes! Blitzy excels at working with existing codebases, using them as a foundation to ensure consistent, high-quality development. The platform enables you to add new features to existing products, generate comprehensive documentation, and tackle technical debt by upgrading legacy systems to state-of-the-art technologies or refactoring complex codebases. Our platform deploys dedicated AI agents that map and understand your codebase before generation, ensuring intelligent, contextualized development that aligns with your existing patterns and standards. What programming languages does Blitzy support? Blitzy's AI platform works with all programming languages. How should I structure my prompts for Blitzy? Structure and organization are crucial when prompting Blitzy. The most effective prompts follow our prompting template with clear sections for WHY (vision & purpose), WHAT (core requirements), and HOW (technical details, user experience & implementation priorities). Each section should be detailed but concise, focusing on essential information while providing relevant context. Including structured frameworks and concrete examples - like data models, user stories, or feature templates - helps Blitzy deliver more precise and purposeful solutions. What information does Blitzy need to compile and run my code? During code generation, Blitzy compiles your codebase and performs runtime validation to ensure the generated code works correctly. To enable this, we require: (1) Internal dependencies - any private packages, libraries, or binaries not publicly available that your code needs to build and run, (2) Environment variables and secrets - API keys, credentials, and configuration values required for compilation and runtime (shared securely through our encrypted UI, never exposed to AI agents), and (3) Build instructions - the specific steps or scripts needed to compile your code, typically found in your README or setup documentation. This information allows Blitzy to replicate your development environment and verify that all generated code functions properly before delivery. How can I exclude certain files or folders from Blitzy's code generation? Create a .blitzyignore file in your repository's root directory to specify which files or paths Blitzy should exclude during tech-spec generation and code generation. This works similarly to .gitignore - simply list the file patterns, directories, or specific files you want Blitzy to skip, using standard gitignore syntax like *.log, /build/, or config/secrets.json. To ensure Blitzy respects these exclusions, mention in both your codebase context prompt and target state prompt that Blitzy should reference the .blitzyignore file and exclude those paths from processing. Can I cancel my project/job (code gen) once in progress? At this time, jobs are not cancelable. Once you submit, it consumes the assigned quota. Build enterprise software in days, not months. Build enterprise software in days, not months. Start building Talk to an expert Build enterprise software in days, not months. Start building Talk to an expert Blitzy One Kendall Square, Cambridge, MA 02139 © 2026 Blitzy. All rights reserved Product Overview Use-cases Security Pricing Company About us Careers Support Help Service status Trust center Resources Docs Blog Videos Social YouTube LinkedIn Legal Website Privacy Policy Platform Privacy Policy Terms of Use Sandbox Terms of Use Blitzy One Kendall Square, Cambridge, MA 02139 © 2026 Blitzy. All rights reserved Product Overview Use-cases Security Pricing Company About us Careers Support Help Service status Trust center Resources Docs Blog Videos Social YouTube LinkedIn Legal Website Privacy Policy Platform Privacy Policy Terms of Use Sandbox Terms of Use Blitzy One Kendall Square, Cambridge, MA 02139 © 2026 Blitzy. All rights reserved Product Overview Use-cases Security Pricing Company About us Careers Support Help Service status Trust center Resources Docs Blog Videos Social YouTube LinkedIn Legal Website Privacy Policy Platform Privacy Policy Terms of Use Sandbox Terms of Use