Nativeclaw

NativeClaw

NativeClaw is a personal AI agent that runs as a managed background service on macOS, Linux, or Windows and talks to you through Telegram.

It started as a Claude Code bridge. Current NativeClaw can run Claude, Codex (OpenAI CLI), or any configured OpenRouter profile as the active backend, keep persistent workspace memory, run scheduled jobs, handle common media inputs, and survive service restarts.

[!WARNING] **Deprecated as of 2026-06-15.** The Telegram bridge is being sunset. Starting June 15, 2026, Claude Agent SDK / `claude -p` usage (which the Claude backend depends on) moves off plan limits onto separate metered credits, changing the cost model this project was built around. **`v2.1.0` is the final feature release.** Existing installs keep running and the Codex / OpenRouter backends are unaffected, but no further bridge development is planned.

What It Does

• Responds to Telegram messages through Claude Code, Codex CLI, or OpenCode for OpenRouter profiles
• Switches backends with /claude, /codex, /or profile , plus convenience aliases such as /kimi, /minimax, and /grok
• Preserves backend-switch continuity with raw gap transcripts
• Auto-compacts long sessions by backend/profile threshold; /compact triggers manual compaction
• Shows live context-window usage per backend with /stats
• Runs bridge-level scheduled jobs such as briefs, audits, queue recovery, and heartbeats
• Keeps durable memory in workspace files instead of relying on chat history
• Supports images, voice notes, audio files, and document attachments
• Uses launchd, systemd, or Windows Task Scheduler so the bridge survives reboot/crash cycles
• Ships a visual setup wizard and an on-demand control panel at 127.0.0.1:9292 (no terminal required after install)

What's Current

Current bridge version: `v2.1.0` (final feature release — see deprecation notice above).

**v2.0 is a major release** — NativeClaw becomes an installable assistant with a visua