Oven

A terminal coding agent. Oven connects to an LLM, reads and writes files in your project, runs shell commands, and helps you get things done.

Warning

Status: not production-ready.

Install

Prebuilt binaries are attached to every release tag. Grab the installer for your platform — it downloads the matching archive, puts oven in ~/.oven/bin (%USERPROFILE%\.oven\bin on Windows), and adds that directory to your PATH.

Linux / macOS — one-liner (latest release):

curl -fsSL https://raw.githubusercontent.com/guuzaa/oven/master/scripts/install.sh | bash

Or pin a specific version:

curl -fsSL https://raw.githubusercontent.com/guuzaa/oven/master/scripts/install.sh | bash -s v0.0.1

Or run the installer from a checkout:

./scripts/install.sh            # latest release
./scripts/install.sh v0.0.1     # or a specific tag

Windows (x86_64) — PowerShell one-liner (latest release):

irm https://raw.githubusercontent.com/guuzaa/oven/master/scripts/install.ps1 | iex

Pin a specific version:

$env:OVEN_VERSION='v0.0.1'; irm https://raw.githubusercontent.com/guuzaa/oven/master/scripts/install.ps1 | iex

Or run the installer from a checkout:

powershell -ExecutionPolicy Bypass -File scripts\install.ps1 v0.0.1

The installer defaults to the latest release; pass a tag argument or set the OVEN_VERSION environment variable to pin one (the v prefix is optional). Restart your terminal after installing, then verify with oven --help. Prebuilt binaries cover Linux x86_64/arm64 (musl), macOS x86_64/arm64, and Windows x86_64.

Usage

# One-shot query
oven "what does this project do?"
oven -Q "what is love?"

# Interactive TUI
oven

# Pick model / API key
OVEN_MODEL=deepseek-v4-flash OVEN_API_KEY=sk-xxx oven

# Resume a session
oven --session my-session
oven -c                 # resume the most recent session in this directory

# Work in a different directory
oven -C /path/to/project

Run oven --help for the full CLI.

Logs

Oven writes a rotating log to ~/.oven/logs/oven.log (10 MiB per file, one backup). The TUI never prints logs to the terminal. Increase verbosity with OVEN_LOG=debug (or RUST_LOG).

How it works

Oven runs inside your project directory and speaks the OpenAI chat completions format, so any compatible provider works (OpenAI gpt-*, DeepSeek deepseek-*, Zhipu glm-*, Kimi kimi-*). The model can read and write files and run shell commands; responses stream back in real time.

Configuration

Config lives in .oven.toml at the project root, or globally at ~/.oven/config.toml (created as a template on first run). Env vars OVEN_MODEL, OVEN_API_KEY, and OVEN_BASE_URL override it.

tools = ["file_read", "file_write", "bash"]

# Fraction of the model's context window that triggers automatic history
# compaction after a turn completes. 0 disables auto-compaction; /compact
# still works. Ignored when the active model's window size is unknown.
compact_threshold = 0.8

[provider]
name = "deepseek"

[providers.deepseek]
model = "deepseek-v4-flash"
base_url = "https://api.deepseek.com"
api_key = "sk-xxx"
reasoning_effort = "high"

[providers.xai]
api_key = "xai-xxx"
reasoning_effort = "low"

• tools — capabilities the agent can invoke (file_read, file_write, bash, glob, grep); an empty list means the defaults.
• [mcps] — MCP servers: stdio (command/args/env) or remote streamable HTTP (url/headers); their tools are mounted as <server>_<tool>.
• AGENTS.md / CLAUDE.md from ~/.oven/ and the project root are injected into the system prompt.
• Skills live in SKILL.md directories under ~/.oven/skills/ and .oven/skills/ (project wins).
• Sessions are JSONL files under ~/.oven/sessions/.

Interactive mode

Type a prompt and press Enter to send. While running, Esc cancels; when idle, Esc rewinds the last exchange and restores your message. Ctrl-C quits, Shift+Tab toggles plan mode. The TUI shows the current todo list, model, and working directory; mouse selection copies automatically. When the context window fills up, history is compacted into a summary automatically (or on demand via /compact). Thinking blocks and tool results collapse to one line — double-click to expand or collapse them, and each turn ends with a Worked for Xs separator. After exiting, the session id is printed so you can resume with --session <id>.

Slash commands

Command	What it does
/clear	New chat (new session, old file kept)
/exit	Quit
/model	Switch model: /model <id> [none|low|medium|high]
/setup	Configure provider: /setup name=... api_key=...
/plan	Toggle plan mode: /plan [on|off]
/compact	Compact conversation history into a summary; auto-compaction triggers at compact_threshold.

Build from source

cargo build -r
./target/release/oven

License

MIT