kbot

name	kbot
description	Open-source terminal AI agent with 11 specialist agents, 214 tools, and 19 providers.
version	2.13.1
license	MIT
metadata	{"kbot":{"requires":{"bins":["node"]},"emoji":"🤖","homepage":"https://kernel.chat","os":["macos","linux"],"install":[{"type":"node","package":"@kernel.chat/kbot","global":true}]}}

kbot

Open-source terminal AI agent by kernel.chat group. Bring your own key, pick your model, run locally.

Install

npm install -g @kernel.chat/kbot

Quick Start

# Interactive mode
kbot

# One-shot
kbot "fix the auth bug in login.ts"

# Force a specialist
kbot --agent researcher "compare React vs Svelte"

# Local models (free, private)
kbot local

Features

• 11 specialist agents (5 built-in + 6 presets) with intent-based routing (kernel, researcher, coder, writer, analyst, and more)
• 214 built-in tools (file ops, bash, git, GitHub, web search, browser automation, Docker sandbox, E2B cloud sandbox, LSP intelligence, linting, testing, sub-agents)
• 19 AI providers (Anthropic, OpenAI, Google, Mistral, xAI, DeepSeek, Groq, Together, Fireworks, Perplexity, Cohere, NVIDIA, SambaNova, Cerebras, OpenRouter, Ollama, LM Studio, Jan, kbot Local)
• Learning engine that adapts to your coding patterns, solutions, and preferences
• Graph memory — knowledge graph with entity extraction and relationship reasoning
• Architect mode — dual-agent plan-review-implement loop for complex refactors
• Provider fallback — auto-failover across 19 providers when one goes down
• Self-evaluation — responses scored for quality and auto-retried if low
• Repo map — Aider-style codebase indexing for automatic context injection
• MCP server for IDE integration (VS Code, Cursor, Windsurf, Zed, Neovim)
• MCP-native plugins — extend kbot with MCP server plugins
• LSP integration — go-to-definition, find-references, hover, rename via language servers
• HTTP server mode for REST API access to all tools
• Local-first with Ollama, LM Studio, Jan, and kbot Local support ($0 cost)
• Mimic profiles to adopt coding styles (claude-code, cursor, copilot, etc.)
• Persistent memory across sessions
• Autonomous planning with Magentic-One-style task ledger
• Confidence calibration — agents know what they know and don't know
• Agent handoff protocol — structured task delegation between specialists
• Shared blackboard — working memory visible to all agents in a session
• Temporal reasoning — checkpoints, backtracking, anticipation
• Abductive & counterfactual reasoning — hypothesis generation and what-if analysis
• Meta-planning — agents evaluate and adapt their own strategies
• Intentionality — quality drives, outcome preferences, intrinsic motivation

MCP Integration

Add kbot as an MCP server in your IDE:

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

Exposes 14 tools: 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 214 tools for any LLM or automation pipeline.

Links

• Live companion: kernel.chat
• npm: @kernel.chat/kbot
• GitHub: isaacsight/kernel