Installation
Heddle is packaged as a global Node.js CLI. Works on macOS, Linux, and Windows with Node.js 20+.
Before you install
- Use Node.js 20 or newer.
- Install with npm globally, or run on demand with npx.
- Configure at least one provider: OpenAI account sign-in, OpenAI API key, Anthropic API key, or a local Ollama server.
- Run Heddle from the repository you want it to inspect, edit, and verify.
Global Install
npm install -g @roackb2/heddleRun without global install
npx @roackb2/heddleProvider Authentication
OpenAI — Account Sign-in (experimental)
heddle auth login openaiUses your ChatGPT / Codex account. Not official OpenAI support — use is subject to OpenAI's terms. Check or remove the credential with heddle auth status and heddle auth logout openai.
OpenAI — API Key
export OPENAI_API_KEY="your_key_here"If both OAuth credential and API key are present, prefer the key with heddle --prefer-api-key.
Anthropic — API Key
export ANTHROPIC_API_KEY="your_key_here"Local Ollama
ollama list
heddle --model ollama/llama3.2:latest ask "Reply with exactly: ok"Start Ollama locally, then choose an installed chat model with the ollama/ prefix. Local models can vary in tool-calling quality, so review traces and diffs carefully.
Choose an auth path
Verify installation
heddle --version
heddle auth status
heddle ask "Reply with OK"The first agent run creates workspace-local state under .heddle/, including saved sessions, traces, memory catalogs, uploads, heartbeat tasks, and run records.
Quick Start
# 1. Navigate to any repository
cd /path/to/your/project
# 2. Start interactive chat
heddle
# 3. Try a prompt
# "Summarize this repo and show the main entrypoints."
# 4. Start the browser control plane (optional)
heddle daemonOptional: local models
If you use Ollama, Heddle can run without a hosted provider API key. The terminal and browser model selectors can show installed Ollama chat models when Ollama is running.
ollama list
heddle --model ollama/llama3.2:latest ask "Reply with exactly: ok"
heddle chat --model ollama/llama3.2:latestOptional: Semantic Drift Telemetry
Install cyberloop in the same environment to enable drift-level observability in chat sessions:
npm install -g cyberloop