{"slug": "agentty-a-drop-in-alternative-to-claude-code-written-in-c-26-11-0-mb-binary", "title": "Agentty – A drop-in alternative to claude-code, written in C++26. 11.0 MB binary", "summary": "Agentty, a drop-in alternative to Claude Code written in C++26, has been released as a single 11.0 MB static binary with sub-millisecond startup. The tool supports multiple AI models including Claude, GPT, Groq, and Ollama, and runs in air-gapped environments via SSH relay. It offers sandboxed execution, memory persistence, and integration with Zed through the Agent Client Protocol.", "body_md": "**AI pair programming in your terminal**\n\nOne static binary. Sub-millisecond startup. Any model.\n\n```\ncurl -fsSL https://raw.githubusercontent.com/1ay1/agentty/master/install.sh | sh\ncd your-project\nagentty\n```\n\nFirst launch opens auth — OAuth (uses your Claude Pro/Max subscription) or paste an API key. Once you're in, a first-run welcome card suggests a few things to try; just type and hit Enter.\n\n|\nCold start under 1ms. No Node, no Python, no npm install. Just a static binary. Claude, GPT, Groq, OpenRouter, Ollama, or any OpenAI-compatible endpoint. Switch live with Every shell call runs inside bwrap (Linux) / sandbox-exec (macOS). File tools refuse paths outside your workspace. |\nRun on a box with no internet. Your laptop relays the bytes over SSH with TLS pinned end-to-end. read · write · edit · bash · grep · glob · git · web · search_docs · task — each with a purpose-built widget. Agent Skills + remember/forget memory. Teach it once, every session knows your conventions. |\n\n```\nagentty                                    # Claude (default)\nagentty --provider openai -m gpt-4o        # GPT\nagentty --provider groq -m llama-3.3-70b   # Groq\nagentty --provider ollama -m qwen2.5-coder # Local model\nagentty --provider openrouter              # Any model via OpenRouter\n```\n\n`--provider`\n\npersists. Switch live in-app with `^P`\n\n.\n\n| Key | Action | Key | Action |\n|---|---|---|---|\n`Enter` |\nSend | `^K` |\nCommand palette |\n`Esc` |\nCancel / reject | `^J` |\nThread list |\n`S-Tab` |\nCycle profile | `^P` |\nModel picker |\n`Alt+Enter` |\nNewline | `^N` |\nNew thread |\n`^G` |\nRun code block | `^←/→` or `Alt+←/→` |\nCycle threads |\n\n**Installation options**\n\n**Linux**\n\n```\n# Debian / Ubuntu\ncurl -fsSLO https://github.com/1ay1/agentty/releases/latest/download/agentty_amd64.deb\nsudo dpkg -i agentty_amd64.deb\n\n# Fedora / RHEL / CentOS\nsudo dnf install https://github.com/1ay1/agentty/releases/latest/download/agentty-x86_64.rpm\n\n# openSUSE\nsudo zypper install https://github.com/1ay1/agentty/releases/latest/download/agentty-x86_64.rpm\n\n# Arch (AUR)\nyay -S agentty-bin\n\n# Alpine\ncurl -fsSLO https://github.com/1ay1/agentty/releases/latest/download/agentty-x86_64.apk\nsudo apk add --allow-untrusted agentty-x86_64.apk\n```\n\n**macOS**\n\n```\nbrew tap 1ay1/tap && brew install agentty\n```\n\n**Windows**\n\n```\nscoop bucket add 1ay1 https://github.com/1ay1/scoop-bucket; scoop install agentty\n# or\nwinget install agentty.agentty\n```\n\n**Anywhere (no package manager)**\n\n```\ncurl -fsSL https://raw.githubusercontent.com/1ay1/agentty/master/install.sh | sh\n```\n\n**From source** (needs a C++26 toolchain — GCC 14+ / recent Clang / MSVC)\n\n```\ngit clone --recursive git@github.com:1ay1/agentty.git\ncd agentty && cmake -B build && cmake --build build -j\n```\n\nAll binaries are a single fully-static executable (x86_64 + aarch64 on Linux, Intel + Apple Silicon on macOS). Packaging details: [ packaging/README.md](/1ay1/agentty/blob/master/packaging/README.md).\n\n**Air-gapped hosts**\n\n```\nagentty airgap --setup user@host   # first time: copies credentials\nagentty airgap user@host           # every time after\n```\n\nYour laptop relays via SOCKS5-over-SSH. TLS pins on real upstreams — the network in between can't MITM you.\n\n**Inside Zed (ACP)**\n\nagentty speaks the [Agent Client Protocol](https://agentclientprotocol.com) — the same protocol Zed uses for Claude Code. Add to Zed's settings:\n\n```\n{\n  \"agent_servers\": {\n    \"agentty\": {\n      \"command\": \"agentty\",\n      \"args\": [\"acp\"]\n    }\n  }\n}\n```\n\n**Run code blocks from replies (Ctrl+G)**\n\nThe AI hands you a fenced block of commands — don't copy-paste it. `^G`\n\nlists\nthe blocks from the last reply; `Enter`\n\n(or a digit) runs one **interactively\non your real terminal**: the TUI suspends, sudo password prompts work, output\nstreams live, `Ctrl+C`\n\nkills the command (not agentty). When it exits, a\nresult card lets you attach the captured output to the composer as a\ncollapsed chip (`a`\n\n), copy it (`y`\n\n), or discard (`Esc`\n\n) — so \"it failed with\nX\" reaches the model without you re-typing anything.\n\nRuns the right shell per block on every OS: `sh`\n\n/`bash`\n\nblocks through\n`/bin/sh`\n\non Linux/macOS, `powershell`\n\n/`pwsh`\n\nand `cmd`\n\n/`bat`\n\nblocks through\nPowerShell / `cmd.exe`\n\non Windows. Prompt `$ `\n\nmarkers are stripped, a block\nyour platform can't run offers edit/copy instead, and capture is capped at\n2 MB. Details: `docs/RUN_CODE_BLOCK.md`\n\n**Agent Skills**\n\nDrop a `SKILL.md`\n\nanywhere under `.agentty/skills/`\n\nor `~/.agentty/skills/`\n\n— it's live next turn. Compatible with Claude Code's `.claude/skills/`\n\nformat.\n\nOn codebases with internal DSLs or tribal conventions, agent accuracy jumps from ~20% to ~85% with curated skills ([research](https://arxiv.org/abs/2410.03981)).\n\n**Architecture**\n\nPure-functional update loop: `(Model, Msg) -> (Model, Cmd)`\n\n. View is `Model -> Element`\n\n, rendered by [maya](https://github.com/1ay1/maya). Process management via `posix_spawn`\n\n+ `poll(2)`\n\n. File writes are atomic (`write`\n\n+ `fsync`\n\n+ `rename`\n\n).\n\nDeep dive: [ docs/ARCHITECTURE.md](/1ay1/agentty/blob/master/docs/ARCHITECTURE.md) ·\n\n`docs/RENDERING.md`\n\n**Releasing (maintainers)**\n\nCutting a release is one command:\n\n```\nscripts/cut-release.sh X.Y.Z      # POSIX / macOS / Linux / Git-Bash\nscripts\\cut-release.cmd X.Y.Z     # Windows cmd.exe\n```\n\nIt bumps `project(agentty VERSION …)`\n\nin `CMakeLists.txt`\n\n(the single source\nof truth every manifest derives from), promotes `CHANGELOG.md`\n\n's `[Unreleased]`\n\nsection to `[X.Y.Z]`\n\n, commits, tags `vX.Y.Z`\n\n, and pushes. The tag push fires\nGitHub Actions, which builds every binary + OS package (Linux x86_64/aarch64\non native runners, macOS Intel/ARM, Windows `.exe`\n\n/`.msi`\n\n) and auto-submits to\nwinget, Homebrew, Scoop, and the AUR — nix/snap/gentoo manifests are attached\nto the release. `--dry-run`\n\npreviews without writing anything.\n\nMIT — see [LICENSE](/1ay1/agentty/blob/master/LICENSE).", "url": "https://wpnews.pro/news/agentty-a-drop-in-alternative-to-claude-code-written-in-c-26-11-0-mb-binary", "canonical_source": "https://github.com/1ay1/agentty", "published_at": "2026-07-15 22:30:24+00:00", "updated_at": "2026-07-15 22:55:20.386582+00:00", "lang": "en", "topics": ["ai-tools", "developer-tools", "ai-agents", "large-language-models"], "entities": ["Agentty", "Claude", "GPT", "Groq", "OpenRouter", "Ollama", "Zed"], "alternates": {"html": "https://wpnews.pro/news/agentty-a-drop-in-alternative-to-claude-code-written-in-c-26-11-0-mb-binary", "markdown": "https://wpnews.pro/news/agentty-a-drop-in-alternative-to-claude-code-written-in-c-26-11-0-mb-binary.md", "text": "https://wpnews.pro/news/agentty-a-drop-in-alternative-to-claude-code-written-in-c-26-11-0-mb-binary.txt", "jsonld": "https://wpnews.pro/news/agentty-a-drop-in-alternative-to-claude-code-written-in-c-26-11-0-mb-binary.jsonld"}}