K:BOT
Open-source terminal AI agent. 37 specialists, 85 tools, 19 providers, local-first.

npx @kernel.chat/kbot

Why K:BOT?

Most terminal AI agents lock you into one provider, one model, one way of working. K:BOT doesn't.

• 19 providers, zero lock-in — Claude, GPT, Gemini, Mistral, Grok, DeepSeek, Groq, SambaNova, Cerebras, OpenRouter, and more. Switch with one command.
• Runs fully offline — Ollama, LM Studio, or Jan for $0 local AI. No data leaves your machine.
• Learns your patterns — remembers what worked, adapts to your coding style, gets faster over time.
• 37 specialist agents — auto-routes your request to the right expert (coder, researcher, debugger, physicist, economist, and 32 more).
• MCP server built in — plug kbot into Claude Code, Cursor, VS Code, Zed, or Neovim as a tool provider.

How it compares

	K:BOT	Claude Code	Aider	OpenCode
AI providers	14	1	6	75+
Specialist agents	37	0	0	0
Learning engine	Yes	No	No	No
Offline mode	Ollama + OpenClaw	No	Ollama	Ollama
MCP server	Yes	N/A	No	No
Web companion	kernel.chat	No	No	No
Open source	MIT	Source available	Apache 2.0	MIT
Cost	BYOK / $0 local	$20+/mo	BYOK	BYOK

Quick Start

# Install globally
npm install -g @kernel.chat/kbot

# Or run directly (no install)
npx @kernel.chat/kbot

# Or use the install script
curl -fsSL https://kernel.chat/install.sh | bash

# Interactive mode
kbot

# One-shot
kbot "explain this codebase"
kbot "fix the auth bug in src/auth.ts"
kbot "create a Dockerfile for this project"

# Pipe mode (for scripting)
kbot -p "generate a migration for user roles" > migration.sql

# Use local models (free, no API key)
kbot ollama

# Set up your API key
kbot auth

Specialists

K:BOT auto-routes to the right agent for each task. Or pick one manually with --agent <name>.

Core: kernel, researcher, coder, writer, analyst Extended: aesthete, guardian, curator, strategist, infrastructure, quant, investigator, oracle, chronist, sage, communicator, adapter Domain: physicist, mathematician, biologist, economist, psychologist, engineer, medic, linguist, ethicist, educator, diplomat Systems: session, scholar, auditor, benchmarker, synthesizer, debugger

Features

• 60+ Tools — File read/write, bash, git, GitHub, web search, Jupyter notebooks, Docker sandbox, browser automation, background tasks, MCP client
• Local-First — Simple tasks (file reads, git, grep) run locally without an API call ($0)
• Learning Engine — Caches successful patterns, solutions, and your preferences
• Mimic Matrix — Code like Claude Code, Cursor, Copilot, or framework experts (Next.js, React, Rust, Python)
• Autonomous Planner — Multi-step tasks get broken into plans, executed, and verified
• Subagent System — Spawn parallel workers for research, coding, and analysis
• Sessions — Save, resume, and share conversations
• Hooks & Plugins — Pre/post tool hooks and custom plugin system
• IDE Integration — MCP server for VS Code, Cursor, Windsurf, Zed, Neovim

Slash Commands

Command	What it does
/agent <name>	Switch specialist agent
/model <name>	Switch AI model
/mimic <profile>	Code like Claude Code, Cursor, Next.js, etc.
/plan <task>	Autonomous plan + execute mode
/save	Save your conversation
/resume <id>	Pick up where you left off
/ollama [model]	Switch to local models
/thinking	Toggle extended thinking
/compact	Compress conversation history
/matrix	Manage custom agents
/plugins	Manage plugins
/help	Full command list

Providers

Provider	Cost	How to set up
Anthropic (Claude)	$3-15/M tokens	ANTHROPIC_API_KEY
OpenAI (GPT)	$2.50-10/M tokens	OPENAI_API_KEY
Google (Gemini)	$0.15-0.60/M tokens	GOOGLE_API_KEY
Mistral	$0.25-2/M tokens	MISTRAL_API_KEY
xAI (Grok)	$3-15/M tokens	XAI_API_KEY
DeepSeek	$0.14-2.19/M tokens	DEEPSEEK_API_KEY
Groq	$0.05-0.27/M tokens	GROQ_API_KEY
Together AI	$0.20-1.20/M tokens	TOGETHER_API_KEY
Fireworks	$0.20-0.90/M tokens	FIREWORKS_API_KEY
Perplexity	$0.20-1.00/M tokens	PERPLEXITY_API_KEY
Cohere	$0.50-15/M tokens	COHERE_API_KEY
NVIDIA NIM	$0.10-0.40/M tokens	NVIDIA_API_KEY
Ollama (Local)	Free	ollama serve
OpenClaw (Local)	Free	openclaw-cmd start

Set any provider's env var and K:BOT auto-detects it. Or run kbot auth for interactive setup.

Architecture

graph TD
    A[User Message] --> B{Local-first check}
    B -->|file, git, grep| C[Instant response — $0]
    B -->|Needs AI| D{Complexity detection}
    D -->|Simple| E[Agent Router → Specialist]
    D -->|Multi-step| F[Autonomous Planner]
    E --> G[Provider API + Tool Loop]
    F --> G
    G --> H{Tools}
    H --> I[File ops, bash, git, GitHub, search, browser...]
    G --> J[Learning Engine]
    J --> K[Patterns + Solutions + User Profile]

Loading

MCP Server

Use kbot as a tool provider inside any MCP-compatible IDE:

{
  "mcp": {
    "servers": {
      "kbot": { "command": "kbot", "args": ["ide", "mcp"] }
    }
  }
}

14 tools exposed: kbot_chat, kbot_plan, kbot_bash, kbot_read_file, kbot_edit_file, kbot_write_file, kbot_search, kbot_github, kbot_glob, kbot_grep, kbot_agent, kbot_remember, kbot_diagnostics, kbot_status.

HTTP Server

kbot serve --port 7437 --token mysecret

REST API exposing all 85 tools. Any LLM or automation pipeline that can make HTTP calls can use kbot as a backend.

Security

• API keys encrypted at rest (AES-256-CBC)
• Config file restricted to owner (chmod 600)
• Destructive operations require confirmation
• Bash tool blocks dangerous commands (rm -rf /, fork bombs, etc.)
• Tool execution timeout (5 min default)
• Result truncation prevents memory exhaustion (50KB)

Development

cd packages/kbot
npm install
npm run dev          # Run in dev mode (tsx)
npm run build        # Compile TypeScript
npm run test         # Run tests
npm run typecheck    # Type-check only

Web Companion — kernel.chat

K:BOT has a web companion at kernel.chat — same 37 agents, persistent memory, and a visual interface. Free to use (20 messages/day).

The web app source lives in src/ with a Supabase backend in supabase/.

Community

• Web: kernel.chat
• npm: @kernel.chat/kbot
• GitHub: isaacsight/kernel
• Issues: Report a bug

License

MIT — kernel.chat group