v0.1.0 — privacy-first coding agent. One binary, no telemetry, skills-driven.
Built with Ratatui and Crossterm.
Respondami has no permission gates — tool calls execute immediately without asking for approval. The agent has full autonomy to read, write, edit files, and run shell commands. Run it inside a sandbox (container, VM, or bubblewrap) to protect your system.
git clone https://github.com/sorcerersr/respondami.git
cd respondami
cargo build --releaseThe binary is at target/release/respondami.
- Start your LLM server (e.g.
llama-server -m model.gguf) - Run Respondami:
./target/release/respondami - Configure: On first run, Respondami creates
~/.config/respondami/config.jsonwith defaults. Edit it to point to your LLM server. - Send a message: Type in the input area and press
Enter.
For a detailed walkthrough, see docs/quick-start.md.
- Session management — persist conversations as JSONL, resume across sessions
- Automatic compaction — LLM summarization keeps conversations within the context window
- Skills — self-contained capability packages (open standard, agentskills.io)
- Hooks — shell scripts at lifecycle points to inject context, block actions, or run side effects
- RTK integration — automatic command rewriting for safer shell execution
- Project instructions —
AGENTS.mdloaded into the system prompt on every turn
| Document | Description |
|---|---|
| Quick Start | Getting started, key bindings, status bar |
| Configuration | All config options, directory layout |
| Skills | Using and creating skills |
| Hooks | Lifecycle hooks, events, examples |
| Logging | Log file, debug levels, SSE debug capture |
| Security | Threat model, sandboxing recommendations |
| Troubleshooting | Common issues and fixes |
cargo build --release # build
cargo run # run the TUI
cargo clippy --all-targets --all-features # must be clean (0 errors, 0 warnings)
cargo test # all tests must passCopyright (c) Stefan "Sorcerer" Rabmund
This project is licensed under the MIT license (LICENSE or http://opensource.org/licenses/MIT)
