Hone

“Not a chat toy designed to indulge you, but a ruthless defender of your investment discipline.”

Honeclaw is dedicated to being a professional investment assistant that truly understands you.

Why the name Hone:

"Hone" means to sharpen, to refine an edge. And serious investing is fundamentally just that kind of process: it is not about chasing every piece of news, nor reacting emotionally to every rise and fall, but about continuously honing one’s judgment through research, comparison, review, and long-term discipline.

English | 简体中文 | Website: hone-claw.com | Blog: hone-claw.com/blog | 💬 Community: Discord

1. 🦅 Honeclaw (Hone Financial)

Honeclaw (or simply Hone) is an open-source personal investment research assistant written in Rust. Unlike the “chatbots” on the market that are accustomed to agreeing with users, Honeclaw is designed as a co-pilot for investment research that is capable of calm thinking, objective judgment, and disciplined restraint.

It integrates into your daily workflow across multiple platforms, helping you track developments at companies you hold, enforce strict investment discipline, run scheduled monitoring tasks, and counter emotional trading impulses with rational data and logic.

The public product website is now live at hone-claw.com. It introduces Hone from the user-facing angle: what Hone is, how public chat works, how portfolio monitoring and scheduled tasks fit into daily research, and where to find the roadmap, GitHub repo, Bilibili, and YouTube demos.

Architecture: Interactive system architecture (HTML) — after cloning the repo, open this file in a browser locally to view the diagram.

Wiki: Complete repository guide and startup wiki — includes the directory map, runtime layout, install paths, source startup modes, ports, configuration, verification, and troubleshooting.

2. ✨ Key Features

• 🧠 An Absolutely Rational Core: It does not flatter and does not follow blindly. When you make investment decisions, it cross-checks them against data and predefined discipline, identifying flaws in your reasoning.
• 📱 Seamless Cross-Platform Access: Supports Web, iMessage, Lark, Telegram, and Discord, so you can engage with your investment brain anytime, anywhere.
• 🗂️ Company Portraits & Long-term Memory: Hone can continuously accumulate company profiles and event timelines in Markdown, helping you preserve investment mainline, key operating metrics, risks, and major developments as a reusable long-term research asset.
• 📊 Position Monitoring & Discipline: Set your take-profit and stop-loss levels, add-to-position logic, and key indicators to watch, and Hone will monitor the market for you like a cold, vigilant sentinel.
• ⏰ Powerful Scheduled Tasks (Cron Jobs): Supports complex scheduled monitoring tasks, such as pre-market briefings, post-market summaries, and automatic analysis after specific earnings releases.
• ⚡ Rust-powered Extreme Performance: Built entirely in Rust at the core, ensuring millisecond-level responsiveness for messages across multiple platforms with minimal footprint.

Official website: hone-claw.com introduces Hone’s public chat, portfolio tracking, scheduled tasks, long-term company memory, cross-platform notifications, and roadmap.

Company Portraits Dashboard: A centralized UI to manage long-term research memories, sync investment mainline developments from chats, and review your customized company knowledge base.

3. 🏗️ Getting Started

Prerequisites

• Environment: A Unix-like system (macOS or Ubuntu recommended).
• Rust: Toolchain Edition 2021 or newer.

Tech stack

• System core: Rust (Tokio, Axum, SSE)
• Backend: Rust
• Client (desktop): Rust (Tauri)
• Frontend: SolidJS + Tailwind v4

Why Rust: see the first Hone Blog post, Why Hone uses Rust, and why more teams should start using Rust.

Supported channels

• Web Console: Modern browser interface with interactive charts.
• Mac App: Native macOS desktop experience.
• IM Integration: Feishu (Lark), Discord, Telegram, iMessage.

Installation and Launch

For the full startup matrix, directory guide, ports, configuration, and troubleshooting notes, see the Hone Wiki.

Option A. One-line Install (macOS/Linux)

curl -fsSL https://raw.githubusercontent.com/B-M-Capital-Research/honeclaw/main/scripts/install_hone_cli.sh | bash
hone-cli doctor
hone-cli onboard
hone-cli start

Option B. Homebrew (macOS/Linux)

brew install B-M-Capital-Research/honeclaw/honeclaw
hone-cli doctor
hone-cli onboard
hone-cli start

Option C. Development Mode

git clone https://github.com/B-M-Capital-Research/honeclaw.git
cd honeclaw
cargo run -p hone-cli -- start --build

What the first startup does

Source checkouts use the local CLI build path. Release and Homebrew installs use the packaged hone-cli start command. The first source build can take about 10 minutes depending on network and CPU.

1. Build: Compiles hone-cli, hone-console-page, hone-mcp, and enabled channel runtime binaries into the local Cargo target dir.
2. Config: Loads canonical config.yaml and generates data/runtime/effective-config.yaml for backend and channel processes.
3. Bring-up: Starts the backend and enabled channel listeners in the foreground.

Choose the Agent’s inference backend

Run the guided setup before starting, or reopen it later:

hone-cli onboard
hone-cli configure --section agent --section channels --section providers

Pick Hone Cloud, a local runner such as Codex ACP or OpenCode ACP, or a legacy local CLI runner such as Gemini / Codex CLI. OpenRouter credentials should live in llm.providers.openrouter.api_key/api_keys; legacy llm.openrouter.* key fields remain migration fallbacks only.

The next section’s screenshots show the full model and channel setup.

After startup, configure models and channels in the client app settings

Left: desktop home—main chat interface, start talking to Hone right away.    Right: Settings—inference backend (cloud or local) and channels such as Feishu, Discord, Telegram, and iMessage.

4. 🌰 Examples

1. Standard Q&A	2. Discord chat	3. Scheduled briefings

These screenshots are illustrative only—Honeclaw supports many more workflows and setups you can unlock as you go.

CASES_EN.md collects real-world Q&A examples from Hone (single-stock logic, follow-up questions, daily portfolio-aware suggestions, deep dives, scheduled tasks, theme scouting, and macro). They are laid out as a two-column table on GitHub for quick reading.

5. 💡 A Note from the Maintainer

“The market is full of noise, and greed and fear are the investor’s greatest enemies. I hope Honeclaw can become your calmest anchor in the trading market.”

To comply with open-source licensing requirements, a number of professional valuation tools, investment research workflows, and proprietary knowledge bases are not included in this public repository.

These cover areas such as:

• Advanced DCF and relative-valuation models
• Sector-specific deep-research workflows
• Curated investment research knowledge bases (e.g., earnings transcripts, analyst report libraries)

If you are interested in accessing these capabilities, feel free to reach out to us:

1. YouTube: 巴芒投研美股频道
2. BiliBili: 巴芒投资
3. Discord Community

6. 🤝 Contributing

Honeclaw is committed to becoming the most professional open-source personal investment research infrastructure in the open-source community. If you are interested in Rust backend development, large language model prompt engineering, or financial data analysis, you are welcome to submit a PR.

Contributors:

• carlisle0615
• Finn-Fengming

📄 License

This project is open-sourced under the MIT license.

Star History