Skip to content

Open source Grok Bot alternative, Bring your own agent and subscription: message Claude Code, Codex, Gemini CLI, Cursor, Copilot and 40+ AI agents as bots on your own computer, from iPhone, Mac, Windows, Linux or terminal. Agent-agnostic and free. Also a Muse and Dots alternative. Built in Rust.

8.6 KB
Raw Blame

Codync documentation

Current documentation describes the checked-out implementation. Index last checked against the tree: 2026-10-19. Source review, a passing build, local integration tests and production acceptance are different checks; a feature documented here is not a claim that its deployed service has been verified.

Start here

Task Read
Find a module or decide where a file belongs File structure
Understand clients, host, cloud and data ownership Architecture, conversation runtime
Work on the macOS, Linux and Windows desktop app (Electron) Desktop app, desktop styles
Build, install, restart and run checks Development, Codync Dev beside production
Write or refactor code in any module Clean code
Change something users see or do (keep the codync.dev manual current) User docs
Change an iPhone or desktop screen or button UI conventions
Change the composer Composer behavior and checks
Ship app and host updates Updates, client and host compatibility
Plan and build Android Android plan, Android status
Test a phone over Cloudflare instead of LAN Cloudflare testing
Configure dev/main or deploy the services Environments and deployment
Run the host on a Linux server or cloud VM Linux servers
Reproduce or check something on Windows (UTM VM) Windows testing
Set up accounts, approvals or desktop SSH Accounts and SSH
Check connectors against live services Connector verification
Change transport or cryptography Remote protocol and shared vectors
Review an issue or pull request Issue and PR review (the review-issue-pr workflow) and its security check
Patch around an agent adapter's bug, or check whether upstream fixed one ACP adapter workarounds
Look into errors or failed turns users hit Usage analytics, iPhone crash tracing, investigations

Features

Plans

Investigations

Production signals (PostHog errors, failed turns) traced to a root cause, newest first. Dated records: the fixes they name have shipped or are listed as open.

Documentation layout

docs/
├── README.md                  # Navigation and maintenance rules
├── architecture/              # Current boundaries, data flow and file ownership
├── features/                  # Current behavior and implementation entry points
├── design/                    # UI rules, system integration and design write-ups for larger features
│   └── reference/             # Reference screenshots the designs follow
├── guides/                    # Build, configuration, testing and deployment steps
├── reference/                 # Wire contracts, shared executable fixtures, adapter workarounds
│   └── fixtures/
├── plans/                     # Work planned or in progress, not released yet
├── investigations/            # Dated root-cause records of production errors
├── archive/                   # Dated plans/audits; never current instructions
└── screenshots/               # Images the repository README shows

Service-local entry points remain in cloud/README.md and relay/README.md. Repository-wide development rules remain in AGENTS.md and CLAUDE.md.

Maintaining these docs

  1. Describe implemented behavior in the present tense. Label proposals and unverified deployment steps explicitly.
  2. Keep one authoritative explanation per topic; link to it from related pages. Link to source for schemas and public APIs instead of copying entire definitions that drift.
  3. Use lowercase, kebab-case filenames and relative links. Source paths written in backticks are relative to the repository root.
  4. When moving a document, update inbound links, code comments, test fixture imports and CI path filters in the same change.
  5. Check local links and run the affected fixture tests. Keep generated images/build logs under build/ or a temporary directory, outside the documentation tree.
  6. Record verification scope and limitations. Do not turn an old test report or historical audit into a current passing status.

Archived plans and audits retain earlier decisions and review context. Their original dates, line numbers and proposed APIs are historical.