Skip to content

Latest commit

 

History

History
 
 

01-overview

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 

README.md

Application guide

This documentation maps the complete remote.futrx application: how to use every product surface, why it is designed around project-scoped agent computers, who can use each capability, and how requests move through the system.

Remote showing an agent conversation beside the running project application

Choose your path

Reader Start here
Product user User guide
Server operator Deployment and operations
Architect or security reviewer Philosophy, then System overview and the Threat model
Contributor API and realtime, Data and frontend state, then the code maps

Read in this order

Document Covers
User guide Task-oriented walkthroughs, screenshots, daily recipes, troubleshooting, and the complete feature reference
00-philosophy.md Product doctrine, local-resource model, project and provider homes, capability and control envelopes, isolation boundaries, invariants, and hardening path
01-system-overview.md Product surfaces, runtime components, and the main end-to-end flow
02-auth-users-and-access.md Admin setup, Google users, provider login, roles, and project access
03-projects-and-containers.md Project creation, LXD lifecycle, secrets, sharing, limits, and inspection
04-chat-and-agents.md Chats, prompt execution, providers, modes, skills, streaming, fork, and rewind
05-workspace-tools.md Attachments, files, terminal, Git history, and browser IDE
06-scheduled-tasks.md Host scheduler, task state, capability-scoped agent tools, overlap, guardrails, and recovery
06-previews-and-browser.md App discovery, HTTPS preview URLs, element inspection, and Agent Browser
07-data-and-frontend-state.md File-backed persistence, workspace files, entities, and UI state
08-api-and-realtime.md HTTP endpoints, WebSockets, events, and access gates
09-deployment-and-operations.md Install, proxying, base images, updates, recovery, and security hardening

Cross-cutting references

Document Covers
../../ARCHITECTURE.md Top-level architecture: topology, layers, data flow, and trust boundaries
../threat-model.md STRIDE threat model per trust boundary, with mitigations and residual gaps
../known-limitations.md Current scaling, operational, and functional constraints

Feature map

flowchart TD
    App["remote.futrx"]
    App --> Identity["Identity and access"]
    App --> Workspace["AI workspaces"]
    App --> Tools["Workspace tools"]
    App --> Runtime["Container runtime"]
    App --> Ops["Self-hosting operations"]

    Identity --> Admin["Local administrator"]
    Identity --> Users["Invited Google users"]
    Identity --> Sharing["Per-project membership"]

    Workspace --> Providers["Claude, Codex, Kimi, Antigravity"]
    Workspace --> Chat["Streaming chats"]
    Workspace --> Skills["Reusable skills"]
    Workspace --> Schedules["One-time and recurring tasks"]

    Tools --> Files["Files and uploads"]
    Tools --> Terminal["Interactive terminal"]
    Tools --> Git["Git history and restore"]
    Tools --> Browser["IDE and browser views"]

    Runtime --> Projects["One LXD container per project"]
    Runtime --> Previews["Discovered app ports"]
    Runtime --> Secrets["Managed environment values"]

    Ops --> Install["Install and HTTPS"]
    Ops --> Update["App and workspace updates"]
    Ops --> Recovery["Health and network repair"]
Loading

Scope notes

  • The User Guide describes the current UI and calls out source-vs-demo discrepancies rather than promoting features that are no longer implemented.
  • Product screenshots are authentic captures from the July 22, 2026 walkthrough; diagrams explain behavior that a still image cannot prove.
  • The browser UI is Preact, TypeScript, Vite, and Tailwind.
  • The backend is one Go process serving the API, WebSockets, and embedded frontend.
  • Project execution happens in LXD containers; the durable workspace and provider homes live on the host and are bind-mounted into each container.
  • Metadata uses JSON and JSONL files rather than a database.