Open-source real-time AI meeting assistant. Self-hosted. No data leaves your machine.
Quick Start • Features • Skills • How It Works • Issues
-purple?style=flat-square){kind=icon}
You're in a meeting. Someone asks you a question. You freeze.
Most AI assistants wait until the speaker finishes their sentence before suggesting a reply. By then, the silence is already awkward.
ClawLine suggests replies while they're still talking. You always have something to say. No awkward pauses. No "that's a great question" stalling.
- Under 2 seconds from speech to reply on screen — not after the sentence ends, during it
- 14 skill modes — Interview, Sales, Negotiation, Code Review, and more. Each shapes how replies are generated
- Claude Code integration — auto-loads your project context so the AI already knows what you're working on
- 100% self-hosted — audio goes from browser to localhost. Zero cloud. Your meetings are yours
- Free and open source — bring your own API keys. No subscription. No vendor lock-in
Speed — Replies appear in <2 seconds. Not after the sentence finishes. While they're still talking.
Privacy — Self-hosted. Audio goes from your browser to localhost. Zero cloud storage. Your meetings stay yours.
Skills — 14 built-in modes: Interview, Sales, Standup, Code Review, Negotiation, and more. Each one shapes how replies are generated.
Claude Code Integration — Auto-loads project context from your Claude Code history. The AI knows what you've been working on without you explaining it.
Custom Skills — Create your own prompt templates for any meeting type. Saved locally.
One command — pip install -e . and you're running. No Docker. No infra. No account creation.
git clone https://github.com/3aLaee/clawline.git
cd clawline
pip install -e .cp .env.example .envGet free keys:
- Deepgram → deepgram.com ($200 free credit)
- Groq → console.groq.com (free tier)
chrome://extensions→ Developer mode ON- Load unpacked → select the
extension/folder
python -m clawline uiStart capturing from any browser tab. Replies appear in real-time.
Browser Tab Audio
|
| Chrome Extension (tabCapture + AudioWorklet)
| 48kHz → 16kHz resampling, Int16 PCM
v
WebSocket (localhost:9876)
|
v
Deepgram Nova-3 (streaming STT, <300ms)
|
v
Groq LLM (Llama 3.3 70B, <300ms TTFT)
|
v
Terminal UI ── Smart Reply panel
Total latency: speech → reply on screen in under 2 seconds.
Skills shape how ClawLine generates replies. Pick one when starting a session.
| Skill | What it does |
|---|---|
| General | Adapts to any conversation |
| Interview | STAR method answers, confident delivery |
| Sales Call | Objection handling, consultative selling |
| Sales Discovery | Qualifying prospects, uncovering pain |
| Standup | Status updates, blockers, plans |
| Code Review | Technical feedback, trade-off analysis |
| Negotiation | Tactical responses, protect your position |
| Investor Pitch | Metric-first answers, risk handling |
| Customer Support | Empathy + resolution, first-call fix |
| Client Discovery | Scope, budget, requirements |
| Recruiter | Behavioral probing, candidate evaluation |
| Product Research | Non-leading questions, five whys |
| Lecture | Class participation, smart questions |
| Podcast | Story extraction, audience-first Qs |
Create via UI or add to skills.json:
[
{
"name": "Team Lead",
"description": "Leading retros and team syncs",
"system_prompt": "Team meeting. User is the lead. Keep discussions on track. Assign action items."
}
]Homepage
Skill Selection
Context Screen
Live Session
| ClawLine | Cluely | |
|---|---|---|
| Price | Free (bring your own keys) | $20-200/mo |
| Open source | Yes | No |
| Self-hosted | Yes | No |
| Your data | Stays on your machine | Cloud |
| Skills system | 14 built-in + custom | Limited |
| Claude Code integration | Yes | No |
| Latency | <2s | ~3-5s |
| Key | Action |
|---|---|
Ctrl+R |
Regenerate reply |
Ctrl+Y |
Copy reply to clipboard |
Ctrl+X |
Clear reply |
Ctrl+P |
Set/update context |
Esc |
End session / go back |
# Terminal UI (recommended)
python -m clawline ui
# With Claude Code context auto-loaded
python -m clawline ui --cc
# Headless mode (no UI)
python -m clawline run -c "Sprint planning for backend team"
# Transcription only
python -m clawline transcribe- Audio: Chrome extension + AudioWorklet (real-time resampling)
- STT: Deepgram Nova-3 (streaming, <300ms)
- LLM: Groq with Llama 3.3 70B (300 tok/s)
- UI: Textual (Python TUI framework)
- Transport: WebSocket on localhost
Contributions welcome. Open an issue first to discuss changes.
git clone https://github.com/3aLaee/clawline.git
cd clawline
pip install -e .
python -m clawline uiMIT
ClawLine is an independent, open-source project created for educational and research purposes. It is not affiliated with, endorsed by, sponsored by, or in any way officially connected to Anthropic (makers of Claude and Claude Code), Cluely, Deepgram, Groq, OpenClaw or any of their subsidiaries or affiliates.The visual design of this project draws inspiration from the color palette and aesthetic of Claude Code's terminal interface. All trademarks, service marks, trade names, product names, and logos mentioned in this repository are the property of their respective owners and are used solely for identification and reference purposes. Such use does not imply any affiliation, endorsement, or sponsorship.All product comparisons are based on publicly available information at the time of writing and are provided for informational purposes only. No guarantee is made regarding the accuracy, completeness, or timeliness of competitive information. Users should verify current pricing and features independently.This software is provided "as is", without warranty of any kind. The authors and contributors accept no responsibility or liability for any consequences arising from the use of this software, including but not limited to use during meetings, interviews, calls, or any other professional or personal interactions. **Users are solely responsible for ensuring their use of this software complies with all applicable laws, regulations, terms of service, and professional conduct standards in their jurisdiction.**By using this software, you acknowledge that recording or transcribing conversations may be subject to consent laws (such as one-party or two-party consent requirements) depending on your jurisdiction. It is your responsibility to ensure you have obtained all necessary permissions before recording or transcribing any conversation.This project is released under the MIT License.