pingdotgg/t3code
T3code
t3code is a platform for operating AI coding agents through a web interface to develop and maintain software projects. It functions as an orchestrator that manages AI agent lifecycles and communication, providing a system for provisioning isolated remote development environments and managing secure connections to cloud-hosted resources.
The project utilizes an event-sourced backend architecture that processes typed commands into persisted domain events to maintain system state. It employs a communication system based on JSON-RPC over WebSockets to synchronize real-time state and events between clients and servers.
The platform includes capabilities for workspace versioning and diffing, using Git-based worktree isolation to provide dedicated environments for specific conversation threads. It further manages asynchronous tasks through queue-backed background workers and implements handshake-based traffic routing to transition clients to direct environment paths.
The repository also serves as a TypeScript monorepo boilerplate, providing a pre-configured full-stack project structure with shared type contracts and standardized build tooling.
Features
- AI Coding Agent Platforms - Provides a complete platform for operating AI coding agents through a web interface to maintain software projects.
- AI Agent Orchestrators - Coordinates the lifecycle and communication of AI agents to automate software development.
- Agent Runtimes - Coordinates the lifecycle, access modes, and session stability of backend AI agent runtimes.
- Agent Operating Interfaces - Provides a minimal web interface to operate AI coding agents for developing and maintaining software projects.
- Conversation Threads - Maintains durable, threaded histories of messages and session state for AI interactions.
- Remote Connection Managers - Manages the discovery, linking, and secure credential issuance for remote cloud-hosted development environments.
- Remote Development Environments - Provisions and links isolated development environments to cloud accounts with secure routing.
- WebSocket RPC Interfaces - Uses JSON-RPC over WebSockets to stream typed events and synchronize state between client and server.
- Event Sourcing - Employs an event-sourced architecture to process typed commands into persisted domain events for state maintenance.
- Worktree Isolations - Uses Git worktrees to provide filesystem isolation for different AI agent conversation threads.
- LLM Response Streaming - Supports both incremental streaming and complete buffered delivery of AI assistant outputs.
- Development State Snapshots - Captures workspace snapshots at specific intervals to enable state restoration and computation of file diffs.
- Cross-Stack Type Contracts - Defines common schemas and types in a dedicated package to ensure consistency between server and client.
- Workspace Change Summaries - Generates file patches and summaries of changes between checkpoints or within a single turn.
- Task Queues - Utilizes queue-backed workers to process long-running background flows and maintain deterministic synchronization.
- JSON-RPC Interfaces - Implements JSON-RPC interfaces to translate high-level orchestration models into low-level runtime commands.
- Output Stream Buffering - Manages AI assistant output by either streaming incremental updates or delivering complete buffered results to the UI.
- Traffic Routing - Manages connection handshakes and routes traffic to direct paths between clients and environments.
- Traffic Routing Strategies - Coordinates initial connection handshakes before transitioning API traffic to direct client-environment paths.
- Monorepo Architectures - Implements a monorepo architecture for a full-stack TypeScript project with shared type contracts.
- Workspace State Versioning - Captures workspace snapshots at specific intervals to allow restoring previous states.
- Asynchronous Task Queueing - Executes long-running asynchronous tasks and side effects using queue-backed workers to ensure deterministic order.
- Side Effect Handlers - Triggers asynchronous follow-up work and handles runtime signals as side effects of domain events.
- Workspace Snapshotting - Captures snapshots of a codebase and computes file patches between checkpoints to track changes.
- Full-Stack Application Boilerplates - Creates a pre-configured monorepo structure to accelerate the development of scalable web applications.
- WebSocket State Synchronization - Streams typed events and state changes via WebSockets to keep the user interface synchronized with the server.
