tau

Tau is a self-hosted, Git-native platform-as-a-service that eliminates traditional DevOps by treating all infrastructure as code stored in Git. Every commit, merge, or push automatically triggers builds, deployments, and infrastructure provisioning with no API calls and no manual configuration. Tau delivers capabilities comparable to Vercel, Firebase, and Cloudflare in a single open-source platform you run on your own servers or VMs.

Built in Go with a pure-Go WebAssembly runtime (wazero), Tau runs application functions in WASM containers for high performance, portability, and security across edge and cloud environments. The platform includes a full suite of integrated services: HTTP/DNS routing with automatic HTTPS, KV databases, object storage, PubSub messaging, load balancing, and private networking, all managed through a web console or CLI.

Tau includes Dream, a local development tool that spins up a hermetic, production-parity simulation of the entire cloud in-process as isolated universes you can test against without Docker or external infrastructure. Spore Drive provides automated multi-node deployment, pushing the Tau platform to your servers in a single command.

The platform is also designed for AI-native workflows: the Dream tool ships with a built-in MCP (Model Context Protocol) server, enabling autonomous AI coding agents to programmatically create and manage local cloud environments, build functions, and run end-to-end tests as part of an agentic development loop.

What You Get

• Git-Native Infrastructure - Every push to your Git repository automatically triggers builds, deployments, and infrastructure provisioning with no API calls or manual CLI commands needed.
• Dream Local Development - A hermetic simulation framework that spins up the entire distributed cloud platform in-process as isolated universes, providing true production-parity testing without Docker or external services.
• Spore Drive Automated Deployment - A deployment system that provisions and configures Tau across multiple servers or VMs with a single command, handling SSH orchestration and service startup automatically.
• WebAssembly Application Runtime - Functions and applications run in WASM containers via wazero (pure Go, zero CGO), providing portable, high-performance execution across diverse infrastructure.
• Integrated Full-Stack Services - Built-in HTTP/DNS with automatic HTTPS, KV databases, object storage, PubSub messaging, and load balancing, all configured declaratively in Git rather than through separate service APIs.
• MCP Server for AI Agents - The Dream tool ships with a built-in Model Context Protocol server, enabling autonomous AI coding agents to create, inspect, and manage local cloud universes programmatically.
• Geolocation-Aware Peer Routing - The Seer service uses Haversine distance computation for location-optimized peer discovery, enabling distributed deployment with automatic traffic routing to the nearest node.
• Web Console - A browser-based UI for connecting to and visualizing your clouds, creating projects, managing functions, monitoring builds, and running deployed functions interactively.

Common Use Cases

• Self-hosted Vercel/Firebase replacement - A startup deploys a fullstack application with serverless functions, a KV database, and file storage entirely on their own VMs, retaining full data ownership and eliminating per-request cloud fees.
• AI-native application development - An AI coding agent uses the Dream MCP server to programmatically spin up a local cloud universe, deploy a new function, run integration tests, and iterate on failures without human intervention.
• Edge-distributed SaaS platform - A team uses Spore Drive to deploy Tau across multiple geographic regions, with Seer routing user requests to the nearest node for low-latency access to their application.
• Serverless microservices on bare metal - A developer defines WASM functions in Git, commits, and Tau automatically builds and deploys them with HTTP routing, auto-HTTPS, and PubSub event triggers with no Kubernetes or container registry required.
• Distributed systems testing and prototyping - A researcher uses DreamCLI to spin up multi-node P2P networks locally, simulate service failures, and validate distributed system behavior before deploying to real infrastructure.

Under The Hood

Architecture Tau is a highly modular, layered Go monorepo organized around clean service boundaries. The core layer defines pure Go interfaces for each platform service (auth, substrate, seer, tns, patrick, monkey, gateway, and hoarder) with no concrete implementations, enabling true dependency inversion. Concrete implementations live in the services package, while the dream package provides a hermetic simulation environment that instantiates all platform services in-process as isolated universes. Build tags gate optional capabilities including enterprise secrets management, IPFS/web3 storage, and dreaming simulation without polluting core logic. The Substrate service acts as the platform runtime hub, orchestrating HTTP, P2P, database, storage, PubSub, and WebAssembly VM attachment in a well-ordered dependency sequence that ensures reliable startup across distributed nodes.

Tech Stack Tau is written entirely in Go 1.26, leveraging the libp2p ecosystem (go-libp2p, go-libp2p-pubsub, IPFS-lite) for peer-to-peer networking and distributed storage. WebAssembly execution uses wazero, a pure-Go WASM runtime with zero CGO dependencies, making Tau cross-platform without native toolchain requirements. DNS is served by a custom implementation using miekg/dns. GitHub webhooks drive the CI/CD pipeline through the Patrick service, with builds executed by the Monkey service. The Seer service implements geolocation-aware peer discovery using Haversine distance computation. An MCP server is embedded in the Dream tool using the official go-sdk, enabling AI agents to manage local development universes. TypeScript and JavaScript client SDKs are published to npm for the Spore Drive and Taucorder tools.

Code Quality The codebase contains extensive test coverage spanning unit tests, integration tests using the Dream simulation framework (tagged with a dreaming build tag), and end-to-end tests that exercise the full distributed service stack locally. Error handling is consistently explicit throughout, using wrapped errors with contextual messages. Go formatting and import organization are enforced via pre-commit hooks with goimports and staticcheck, and GitHub Actions runs tests against pull requests. Interface-driven design enables clean test doubles across service boundaries. A minor area for improvement is the limited use of typed custom errors, as most error paths use string-wrapped errors, which constrains programmatic error handling in client code.

What Makes It Unique The Dream simulation framework is Tau’s most distinctive engineering contribution: the ability to instantiate the entire distributed cloud platform including multiple P2P nodes, DNS, storage, compute, and messaging as isolated in-process universes for development and testing, using the exact same runtime as production. No containers, no external services, complete production parity. The integration of an MCP server directly in the Dream tool enables AI coding agents to participate in the development loop as first-class actors. The platform’s approach to the Substrate service, treating HTTP routing, P2P messaging, KV storage, object storage, PubSub, and WASM execution as uniformly attached components under one typed interface, creates a uniquely coherent self-contained cloud runtime that eliminates the glue-code overhead of assembling these primitives from separate infrastructure tools.

Self-Hosting

Tau is released under the BSD 3-Clause license, one of the most permissive open-source licenses available. You can use it commercially, modify it, and distribute it freely as long as you retain the copyright notice and license text. There are no copyleft implications: you are not required to open-source applications you deploy on Tau, and you can distribute modified versions of Tau itself without releasing source changes. The codebase includes enterprise build-tag-gated features (secrets management via the ee build tag) but these are included in the open-source repository without a separate commercial license requirement.

Running Tau yourself is operationally substantial. The platform is composed of multiple cooperating services (Substrate, Auth, Seer, TNS, Patrick, Monkey, Gateway, Hoarder) that must start together and maintain P2P connectivity. You are responsible for provisioning servers or VMs, managing TLS certificate renewal via the built-in ACME integration, handling DNS configuration for your domain, maintaining uptime, and performing version upgrades. The Spore Drive tool reduces the initial deployment burden significantly, handling SSH orchestration and firewall configuration, but ongoing operational responsibility including monitoring, backup of distributed KV state, and debugging P2P connectivity issues remains with your team. The Dream framework helps validate changes locally before production, which reduces deployment risk.

There is no managed or hosted version of Tau offered by Taubyte at this time, so there is no paid tier to compare against. What you forego compared to managed platforms like Vercel or Firebase is their SLA guarantees, enterprise support contracts, managed certificate and scaling operations, and global CDN infrastructure. Taubyte provides community support via Discord and documentation at tau.how. Teams adopting Tau should be comfortable with distributed systems operations and plan for the additional engineering overhead of managing a self-hosted multi-service platform.