codebot-ai

CodeBot AI

Zero-dependency autonomous AI agent

Your local-first AI coding agent. Runs with Ollama (no API keys), edits code, browses the web, automates git/PRs/tests. Multi-LLM swarm mode lets multiple AI models collaborate on your tasks. Enterprise security from day 1.

npm install -g codebot-ai && codebot --setup

Works with any LLM — Ollama (local, free), Claude, GPT, Gemini, DeepSeek, Groq, Mistral, Grok.
Web dashboard, VS Code extension, GitHub Action, multi-LLM swarm, and enterprise security built in.

Install & Run

# Install globally
npm install -g codebot-ai

# Launch interactive mode
codebot

# Or run without installing
npx codebot-ai

First run launches the setup wizard — pick your model, paste an API key (or use local LLM), done.

# Local LLM (no API key needed)
ollama pull qwen2.5-coder
codebot --setup       # select "ollama"

# Web dashboard
codebot --dashboard   # opens localhost:3120

# Full autonomous mode
codebot --autonomous "refactor auth module and run tests"

What It Does

Capability	How
Write & edit code	Reads your codebase, makes targeted edits, runs tests
Run commands	Shell execution with security filtering and sandbox support
Browse the web	Controls Chrome via DevTools Protocol — navigate, click, type, screenshot
Search the internet	Real-time web search for docs, APIs, current info
Web dashboard	Sessions, audit trail, metrics, and Command Center at localhost:3120
Schedule routines	Cron-based recurring tasks — monitoring, reports, automation
Persistent memory	Remembers preferences and context across sessions
Enterprise security	RBAC policies, risk scoring, encrypted audit trail, SARIF export
Multi-LLM Swarm	Multiple AI models collaborate using debate, pipeline, fan-out, and more

Multi-LLM Swarm

Launch a swarm of AI agents that collaborate on complex tasks. Mix cloud and local models freely.

codebot --dashboard   # open http://localhost:3120, click "Launch Swarm"

6 Strategies:

• Auto — Router analyzes your task and picks the best strategy
• Debate — Multiple agents propose solutions and vote on the best
• Mixture of Agents — Diverse proposals merged by a synthesizer
• Pipeline — Sequential stages: plan → research → code → review → test
• Fan-Out — Parallel subtasks gathered and synthesized
• Generator-Critic — One agent generates, another critiques, iterate to quality

Supported providers: Anthropic (Claude), OpenAI (GPT), Ollama (local, free), Gemini, DeepSeek, Groq, Mistral, xAI (Grok)

Select providers in the dashboard, type your task, and watch agents stream results in real-time via SSE.

Web Dashboard

Launch with codebot --dashboard or standalone with npx codebot-ai && open http://localhost:3120.

Sessions — Browse and inspect every conversation with message counts and timestamps.

Audit Trail — Cryptographic hash-chained log of every tool execution. One-click chain verification.

Metrics — Session counts, audit events, tool usage breakdown, and activity charts.

Command Center — Interactive terminal, quick actions (git status, run tests, health check), and when connected to the agent: AI chat and tool runner.

8 LLM Providers

Provider	Models
Local (Ollama/LM Studio/vLLM)	qwen2.5-coder, qwen3, deepseek-coder, llama3.x, mistral, phi-4, codellama, starcoder2
Anthropic	claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5
OpenAI	gpt-4o, gpt-4.1, o1, o3, o4-mini
Google	gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash
DeepSeek	deepseek-chat, deepseek-reasoner
Groq	llama-3.3-70b, mixtral-8x7b
Mistral	mistral-large, codestral
xAI	grok-3, grok-3-mini

export ANTHROPIC_API_KEY="sk-ant-..."    # or any provider
codebot --model claude-sonnet-4-6

Or use a local model with zero API keys:

ollama pull qwen2.5-coder && codebot --provider ollama --model qwen2.5-coder

32 Built-in Tools

Tool	Permission	Description
read_file	auto	Read files with line numbers
write_file	prompt	Create or overwrite files (undo snapshots)
edit_file	prompt	Find-and-replace edits with diff preview
batch_edit	prompt	Multi-file atomic find-and-replace
execute	always-ask	Run shell commands (security-filtered)
glob	auto	Find files by pattern
grep	auto	Search file contents with regex
git	prompt	Git operations (status, diff, log, commit, branch)
browser	prompt	Chrome automation via CDP
web_fetch	prompt	HTTP requests and API calls
web_search	prompt	Internet search with summaries
think	auto	Internal reasoning scratchpad
memory	auto	Persistent memory across sessions
routine	prompt	Schedule recurring tasks with cron
code_analysis	auto	Symbol extraction, imports, outline
code_review	auto	Security scanning and complexity analysis
multi_search	auto	Fuzzy search: filenames, content, symbols
task_planner	auto	Hierarchical task tracking
diff_viewer	auto	File comparison and git diffs
test_runner	prompt	Auto-detect and run tests (jest, vitest, pytest, go, cargo)
docker	prompt	Container management (ps, run, build, compose)
database	prompt	Query SQLite databases (blocks destructive SQL)
http_client	prompt	Advanced HTTP with auth and headers
image_info	auto	Image dimensions and metadata
pdf_extract	auto	Extract text and metadata from PDFs
ssh_remote	always-ask	Remote command execution via SSH
notification	prompt	Webhook notifications (Slack, Discord)
package_manager	prompt	Dependency management (npm, yarn, pip, cargo, go)
app_connector	prompt	GitHub, Jira, Linear, Slack integrations
graphics	prompt	Image processing: resize, crop, watermark, convert
delegate	prompt	Multi-agent task delegation

Permission levels: auto = runs silently, prompt = asks first (skipped in --autonomous), always-ask = always confirms.

App Connectors

Connect to external services with OAuth or API keys:

Connector	Capabilities
GitHub	Issues, PRs, repos, code search
Jira	Issues, projects, sprints, transitions
Linear	Issues, projects, teams, cycles
Slack	Messages, channels, users, threads
OpenAI Images	DALL-E generation, editing, variations
Replicate	Run any ML model via API

Credentials stored in encrypted vault (AES-256-GCM).

Security

Built for enterprise from day one:

 Risk Scoring       6-factor risk scoring on every tool call (0-100)
 Encryption         AES-256-GCM encryption at rest
 Audit Trail        SHA-256 hash-chained, tamper-evident logs
 Sandbox            Docker-based execution with network/CPU/memory limits
 RBAC               Declarative JSON policy engine with per-tool permissions
 SARIF Export       GitHub Code Scanning integration
 SSRF Protection    Blocks localhost, private IPs, cloud metadata
 Secret Detection   15+ patterns (AWS keys, tokens, private keys)
 Path Safety        Blocks writes to system directories
 Session Integrity  HMAC-based tamper detection

See SECURITY.md for the full model.

CLI Reference

codebot                                        # Interactive REPL
codebot "fix the bug in app.ts"                # Single task
codebot --autonomous "refactor auth and test"  # Full auto
codebot --continue                             # Resume last session
codebot --dashboard                            # Web dashboard
codebot --tui                                  # Terminal UI (panels)
codebot --doctor                               # Environment health check
echo "explain this error" | codebot            # Pipe mode

All CLI flags

--setup              Run the setup wizard
--model <name>       Model to use
--provider <name>    Provider: openai, anthropic, gemini, deepseek, groq, mistral, xai
--base-url <url>     LLM API base URL
--api-key <key>      API key (or use env vars)
--autonomous         Skip all permission prompts
--resume <id>        Resume a session by ID
--continue, -c       Resume the most recent session
--max-iterations <n> Max agent loop iterations (default: 50)
--tui                Full terminal UI mode
--dashboard          Web dashboard on localhost:3120
--doctor             Environment health checks
--theme <name>       Color theme: dark, light, mono
--no-animate         Disable animations
--no-stream          Disable streaming display
--verbose            Debug output

Interactive commands

/help       Show commands           /model     Show or change model
/models     List supported models   /sessions  List saved sessions
/routines   List routines           /auto      Toggle autonomous mode
/undo       Undo last edit          /usage     Token usage
/clear      Clear conversation      /compact   Force context compaction
/metrics    Session metrics         /risk      Risk assessment history
/config     Show configuration      /doctor    Health checks
/toolcost   Per-tool cost breakdown /rate      Rate limit status
/theme      Switch color theme      /quit      Exit

VS Code Extension

code --install-extension codebot-ai-vscode-2.0.0.vsix

Sidebar chat panel, inline diff preview, status bar (tokens, cost, risk level), theme integration.

GitHub Action

- uses: zanderone1980/codebot-ai/actions/codebot@v2
  with:
    task: review    # or: fix, scan
    api-key: ${{ secrets.ANTHROPIC_API_KEY }}

Programmatic API

import { Agent, AnthropicProvider } from 'codebot-ai';

const agent = new Agent({
  provider: new AnthropicProvider({
    apiKey: process.env.ANTHROPIC_API_KEY,
    model: 'claude-sonnet-4-6',
  }),
  model: 'claude-sonnet-4-6',
  autoApprove: true,
});

for await (const event of agent.run('list all TypeScript files')) {
  if (event.type === 'text') process.stdout.write(event.text || '');
}

Plugins & MCP

Custom tools: Drop .js files in .codebot/plugins/:

module.exports = {
  name: 'my_tool',
  description: 'Does something useful',
  permission: 'prompt',
  parameters: { type: 'object', properties: { input: { type: 'string' } }, required: ['input'] },
  execute: async (args) => `Result: ${args.input}`,
};

MCP servers: Create .codebot/mcp.json:

{
  "servers": [{ "name": "my-server", "command": "npx", "args": ["-y", "@my/mcp-server"] }]
}

Build from Source

git clone https://github.com/zanderone1980/codebot-ai.git
cd codebot-ai
npm install && npm run build
./bin/codebot

Stability

• Auto-retry — exponential backoff on network errors, rate limits, server errors
• Stream recovery — reconnects if the LLM drops mid-response
• Context compaction — smart summarization when hitting context limits
• Process resilience — catches unhandled exceptions, keeps the REPL running
• 1125 tests — comprehensive coverage across agent, tools, security, and dashboard

npm · GitHub · Changelog · Roadmap · Security

MIT — Ascendral Software Development & Innovation