Thunderbolt
AI You Control: Choose your models. Own your data. Eliminate vendor lock-in.
Last verified:
What is Thunderbolt?
Thunderbolt is an open-source, cross-platform AI client that gives enterprises complete control over their AI infrastructure without sacrificing capability. It is a sovereign AI client designed as an extensible workspace where users can interact with AI through chat, search, and research, connect to enterprise data, and choose the models and tools that fit their needs. The tool allows organizations to run AI with their choice of models from leading commercial providers to open-source and local models, connect to systems and data via integrations with pipelines and open protocols, and automate workflows and recurringTasks.
Key features include native applications for web, macOS, Windows, Linux, iOS, and Android; MCP (Model Context Protocol) support for custom integrations; ACP (Agent Client Protocol) compatibility for connecting agents; self-hosted deployment options with optional end-to-end encryption; device-level access controls; reusable workflows for recurring tasks; and native integration with deepset's Haystack platform for RAG and agent orchestration. Thunderbolt is built for enterprise use with Forward-Deployed Engineering support from trusted partners.
Thunderbolt is designed for enterprises and organizations that need data sovereignty, want to eliminate vendor lock-in, and require AI infrastructure that stays entirely within their hands. It targets data sovereignty and vendor lock-in concerns that enterprises raise about offerings from OpenAI, Microsoft, and Anthropic. The tool is suitable for ML engineers, platform teams, government agencies, aerospace manufacturers, and multinational organizations that need to keep sensitive data on-premises.
Thunderbolt pricing
Pricing model: Freemium
Thunderbolt is open-source under MPL 2.0 license and available via waitlist at thunderbolt.io. There is no public free tier mentioned - the source code is freely available on GitHub. For enterprise deployments, pricing reflects support level, customization requirements, and deployment complexity. Integration partners can package Thunderbolt with sovereign storage, infrastructure management, and Forward-Deployed Engineering support as part of comprehensive sovereign AI solutions. Organizations interested in pilot deployments or enterprise licensing should contact [email protected]. Managed-host plans are planned but not yet available.
Thunderbolt pros
- Open-source with MPL 2.0 license for full auditability
- Self-hostable on customer infrastructure anywhere
- Data sovereignty - data never leaves your control
- Cross-platform native apps for web, desktop, and mobile
- Support for any OpenAI-compatible model API
- ACP-compatible agent connectivity
- MCP support for custom system integrations
- Native integration with deepset's Haystack for RAG
- Optional end-to-end encryption for security
- Device-level access controls for enterprise security
- Reusable automation workflows for recurring tasks
- Model and agent agnostic - no vendor lock-in
- On-prem, sovereign cloud, or air-gapped deployment
- Forward-Deployed Engineering support available
- European delivery through trusted partners
- Active development with 4.7k GitHub stars
- Full API for custom integrations
Thunderbolt cons
- Still under active development, not production-ready
- No public inference endpoint - must bring your own model
- Currently depends on authentication (not fully offline-first)
- Search functionality can be disabled but is currently required
- Undergoing security audit before enterprise production readiness
- Requires own AI provider API keys (Anthropic, OpenAI, etc.)
- Self-hosting requires Docker, Bun, Rust toolchain setup
- Limited connector breadth for enterprise systems currently
- Managed-host plans still planned, not yet available
- Security and compliance certifications still in progress
Frequently asked questions about Thunderbolt
What is Thunderbolt?
Thunderbolt is an open-source, cross-platform AI client that gives organizations full ownership of their data, freedom from vendor dependencies, and AI infrastructure that stays entirely within their hands. It is a sovereign AI client designed as an extensible workspace for chat, search, research, automation, and cross-device workflows.
Is Thunderbolt free?
Thunderbolt is open-source under the Mozilla Public License 2.0, and the source code is available on GitHub. However, for enterprise deployments, pricing depends on support level, customization requirements, and deployment complexity. Managed-host plans are planned but not yet available.
What models does Thunderbolt support?
Thunderbolt connects any ACP-compatible agent or any model with an OpenAI-compatible API, including Claude, Codex, OpenClaw, DeepSeek, and OpenCode. It supports frontier, local, and on-prem models. You need to add your own model providers - API keys for Anthropic, OpenAI, Mistral, Fireworks, or any OpenAI-compatible endpoint (Ollama and llama.cpp recommended for local inference).
Can I self-host Thunderbolt?
Yes, Thunderbolt is designed to be self-hosted on your infrastructure. You can deploy your own backend with Docker and sign up to test locally. Self-hosting with Docker Compose or Kubernetes is supported. Your data never leaves your control with on-prem, sovereign cloud, or air-gapped deployment options.
What platforms does Thunderbolt run on?
Thunderbolt has native applications for web, macOS, Windows, Linux, iOS, and Android. It is a cross-platform AI client available on all major desktop and mobile platforms.
What is the difference between Thunderbolt and Haystack?
Thunderbolt is the AI Client Experience layer - a sovereign AI client enabling chat, search, research, automation, and cross-device workflows through a self-hostable interface. Haystack is the AI Infrastructure & Orchestration layer where AI systems are built, deployed, and governed, enabling agents, RAG applications, and data-driven systems. Thunderbolt has native integration with deepset's Haystack for RAG and agent orchestration.
What security features does Thunderbolt offer?
Thunderbolt offers self-hosted deployment, optional end-to-end encryption, and device-level access controls. The project is currently undergoing a security audit and preparing for enterprise production readiness. It enables organizations to maintain security with data staying on their infrastructure.
What is Model Context Protocol (MCP) support?
MCP (Model Context Protocol) support allows Thunderbolt to integrate with your systems through custom integrations. It is one of the open protocols Thunderbolt supports for connecting to enterprise data and pipelines, along with Agent Client Protocol (ACP) for agent frameworks.
Who develops Thunderbolt?
Thunderbolt is developed by MZLA Technologies Corporation, a wholly owned subsidiary of the Mozilla Foundation. MZLA is the organization behind Thunderbird and is guided by principles of openness, user control, and privacy. Thunderbolt is funded through a dedicated investment from Mozilla and is developed by a separate team focused on enterprise AI products.
How do I get started with Thunderbolt?
Thunderbolt is available now via waitlist at thunderbolt.io. For local development, you can clone the repository and run make doctor to verify your tools, make setup to install dependencies, make up to start Postgres and PowerSync in Docker, and make run to start the backend and frontend. For enterprise deployments, contact [email protected] about pilot deployments or enterprise licensing.