Skip to content
🦀 SuperAgent Hermes Agent soul · OpenClaw vision · 11 LLM providers · learns & improves

Your SuperAgent Built in Rust

Personal + coding SuperAgent that grows with you. Powered by the soul of Hermes Agent and the vision of OpenClaw — delivered as a single 15 MB Rust binary that starts in < 50 ms.

npm install -g edgecrab-cli

or: pip install edgecrab-cli · cargo install edgecrab-cli

< 50 ms startup
15 MB binary size
0+ tests
0 LLM providers
Integrates with
GitHub Copilot OpenAI Anthropic Ollama VS Code ACP
Features

Two Great Traditions. One Rust Binary.

EdgeCrab is a SuperAgent — personal assistant and coding agent in one. It inherits Hermes Agent's ethos of deeply personal, steerable AI reasoning and OpenClaw's vision of an always-on assistant living in your tools and channels. Delivered as a single 15 MB static binary: no Python venv, no Node.js, no warmup — 70× lighter than OpenClaw, 1000× faster to start on edge hardware.

Blazing-Fast ratatui TUI

GPU-composited, 60 fps capable terminal UI built with ratatui. Streaming output, slash commands, session management, and full keyboard navigation.

ReAct Tool Loop

Autonomous Reason + Act loop executes file, terminal, web search, memory, process, and skill tools. Recursion-safe with configurable depth limits.

Multi-Provider LLM

GitHub Copilot, OpenAI, Anthropic, Google Gemini, xAI Grok, DeepSeek, Hugging Face, Z.AI, OpenRouter, Ollama, LM Studio — hot-swappable via /model command.

ACP Protocol

Built-in JSON-RPC 2.0 stdio adapter for the Agent Communication Protocol. Registers as a VS Code Copilot agent or any ACP-compatible runner.

Built-in Security

Compiled-in path safety, SSRF protection, command scanning, and output redaction. No runtime patches required — security is a first-class citizen.

Single Static Binary

15 MB binary. < 50 ms cold startup. ~15 MB resident memory. No Python venv, no Node.js, no JVM. Runs anywhere Rust runs.

Python & Node.js SDKs

Async-first Python SDK (edgecrab-sdk on PyPI) and TypeScript-first Node.js SDK (edgecrab-sdk on npm). Both ship with streaming and CLI.

SQLite State & FTS5 Search

WAL-mode SQLite for session persistence. FTS5 full-text search across all session history. Migrate from hermes-agent with a single command.

Three Ways to Install — Pick Yours

npm (no Rust required) · pip (no Rust required) · cargo (compile or grab a release binary). All three install the same native binary. Your Super Powerful Personal Assistant is live in under 90 seconds.

# Install globally via npm — no Rust required
npm install -g edgecrab-cli

# Run setup wizard (detects API keys, writes config)
edgecrab setup

# Verify health
edgecrab doctor

# Start chatting
edgecrab
edgecrab "summarise the git log for today"
# Install via pip — no Rust required
pip install edgecrab-cli

# Run setup wizard (detects API keys, writes config)
edgecrab setup

# Verify health
edgecrab doctor

# Start chatting
edgecrab
edgecrab "summarise the git log for today"
# Install from crates.io
cargo install edgecrab-cli

# Run setup wizard (detects API keys, writes config)
edgecrab setup

# Verify health
edgecrab doctor

# Start chatting
edgecrab
edgecrab "summarise the git log for today"
edgecrab --model openai/gpt-4o "explain this codebase"
pip install edgecrab-sdk

from edgecrab import Agent

agent = Agent(model="anthropic/claude-sonnet-4-20250514")
reply = agent.chat("Explain Rust ownership in 3 sentences")
print(reply)

# Async + streaming
import asyncio
from edgecrab import AsyncAgent

async def main():
  agent = AsyncAgent(model="copilot/gpt-4.1-mini")
  async for token in agent.stream("Write a Rust hello-world"):
      print(token, end="", flush=True)

asyncio.run(main())
npm install edgecrab-sdk

import { Agent } from "edgecrab-sdk";

const agent = new Agent({ model: "openai/gpt-4o" });
const reply = await agent.chat("Review this TypeScript code");
console.log(reply);

// Streaming
for await (const token of agent.stream("Write a README")) {
process.stdout.write(token);
}
# Pull pre-built multi-arch image
docker pull ghcr.io/raphaelmansuy/edgecrab:latest

# Run gateway server
docker run -p 8080:8080 \
-e OPENAI_API_KEY=sk-... \
ghcr.io/raphaelmansuy/edgecrab:latest \
edgecrab gateway --bind 0.0.0.0:8080

# Or use docker-compose
docker compose up
Inspired By

The DNA of EdgeCrab

EdgeCrab is not a fork of either project. It synthesises what made each great — Hermes Agent's soul of autonomous learning + OpenClaw's messaging vision — then compresses it into a Rust binary that runs anywhere.

Feature Hermes Agent 🧠OpenClaw 🦞EdgeCrab 🦀
What it is Autonomous agent (Python, MIT) — 40+ tools, grows with you Personal assistant platform (TypeScript) — 24+ channels SuperAgent — 62 built-in tools, personal + coding, learns & improves (Rust)
Core idea Lives on your server, remembers what it learns, gets more capable over time Always-on assistant living in your channels & tools ✓ Synthesises both: Hermes soul (reasoning + memory) + OpenClaw vision (channels)
User-first alignment ✓ Hermes LLM — steerable & user-aligned; Honcho cross-session user model Partial — depends on system prompt ✓ Compiled-in Honcho user model — deepens with every session
Deep reasoning ✓ Multi-turn + subagent delegation + parallel batch workers Via LLM API calls ✓ ReAct loop + Mixture-of-Agents (4 frontier models in parallel)
Coding agent ✓ Terminal, file, browser & execute_code (Docker / SSH / Modal / Singularity) Partial — via shell/git tools ✓ ReAct + file / terminal / browser / execute_code (Docker, SSH, Modal, local)
Memory & learning ✓ Persistent memory + auto-generated skills + Honcho user model Partial — session memory ✓ SQLite FTS5 sessions + Honcho cross-session user model + skill hub
Always-on presence ✓ 7 platforms: Telegram, Discord, Slack, WhatsApp, Signal, Email, CLI ✓ 24+ channels: WhatsApp, Telegram, Slack, Discord, Matrix, IRC… ✓ 11 gateways: Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Email, Mattermost, SMS, DingTalk, HA + ACP / VS Code
Smart home ✗ Not supported ✗ Not supported ✓ Home Assistant — list entities, get state, list & call services
MCP support ✓ Full MCP client (optional dep — mcp_tool.py, ~1050 lines) ✗ Not documented ✓ Built-in MCP client — tools, resources & prompts (JSON-RPC 2.0)
Install method curl | bash (Python + uv managed) npm install (Node.js required) ✓ npm · pip · cargo — pick yours
Runtime footprint Python runtime · uv managed Node.js runtime · > 1 GB RAM ✓ Rust binary · ~15 MB RAM
Cold startup ~1–3 s (Python + uv) > 500 s on 0.8 GHz edge hardware ✓ < 50 ms everywhere
Security ✓ 5 sandboxing backends (Docker, SSH, Modal, Singularity, local) Runtime patches ✓ Compiled-in (path jail, SSRF, cmd scan, secret redact) + Docker / SSH / Modal backends

From the Soul of Hermes to the Speed of Rust

EdgeCrab is open source, Apache 2.0. Your SuperAgent — personal + coding, grows with you. Install via npm, pip, or cargo in seconds. Runs anywhere on macOS, Linux, and Windows.

npm install -g edgecrab-cli