One binary. 51 commands. 17 agent tools. 29 connectors. Model-agnostic. Composable via pipes. Zero vendor lock-in.
Website · Docs · Commands · Architecture · Contributing
Any AI with exec() becomes an autonomous agent. Every command returns structured JSON. Pipe anything into anything.
jeriko sys | jeriko notify pipe system info to Telegram
jeriko browse --screenshot "url" browser automation
jeriko stripe customers list Stripe API operations
jeriko github prs --create "title" GitHub pull requests
jeriko ai --image "a red fox" DALL-E image generation
jeriko twilio call +1... --say "hi" make a phone call
jeriko vercel deploy --prod deploy to Vercel
jeriko interactive AI chat (REPL)
# One-line install (recommended)
curl -fsSL https://jeriko.ai/install | sh
# From source (requires Bun >= 1.1.0)
git clone https://github.com/etheonai/jeriko.git
cd jeriko && bun install && bun run build
# First-run setup
jeriko initjeriko init configures your AI backend (Claude, OpenAI, Ollama, or any OpenAI-compatible provider), messaging channels, and security settings.
See docs/INSTALL.md for detailed instructions.
jeriko sys # system info (CPU, RAM, disk)
jeriko search "Node.js streams" # web search
jeriko browse --screenshot "https://x.com" # browser automation
jeriko stripe customers list # Stripe API
jeriko github issues --create "Bug" # GitHub issues
jeriko ai --image "a sunset over mountains" # DALL-E image generation
jeriko code --python "print(2**100)" # execute Python
jeriko sys | jeriko notify # pipe output to Telegram
jeriko # interactive AI chat (REPL)Jeriko is model-agnostic with 22 built-in provider presets and automatic env var discovery.
# Anthropic
ANTHROPIC_API_KEY=sk-ant-...
# OpenAI
OPENAI_API_KEY=sk-...
# Local models (Ollama, LM Studio, llama.cpp — runs fully offline)
LOCAL_MODEL_URL=http://localhost:11434/v1
LOCAL_MODEL=llama3.2
# Any OpenAI-compatible or Anthropic-compatible provider
jeriko provider add deepseek --type openai-compatible --base-url https://api.deepseek.com/v1Built-in presets: anthropic, openai, google, deepseek, mistral, groq, together, fireworks, perplexity, cohere, x-ai, cerebras, sambanova, hyperbolic, openrouter, ollama, lm-studio, llama-cpp, jan, text-gen-webui, koboldcpp, vllm.
Organized in 11 categories: system, files, browser, desktop, communication, AI, project scaffolding, payments, cloud, storage, and server management.
Every command outputs structured data: {"ok":true,"data":{...}} or {"ok":false,"error":"...","code":N}. Three formats: json, text, logfmt.
See the full reference: docs/COMMANDS.md
OAuth and API key integrations with: Stripe, GitHub, PayPal, Twilio, Vercel, Google Drive, Gmail, Outlook, OneDrive, HubSpot, Shopify, Slack, Discord, X, Instagram, Threads, Square, GitLab, Notion, Linear, Jira, Airtable, Asana, Mailchimp, Dropbox, SendGrid, Salesforce, Cloudflare.
jeriko connect github # OAuth flow
jeriko connectors # list all with status
jeriko disconnect github # revoke credentialsWhen running as an AI agent, Jeriko provides tools for: shell execution, file I/O, browser automation, web search, screenshots, webcam, connector API calls, sub-agent delegation, skills, persistent memory, parallel tasks, image generation, and web development.
jeriko with no arguments launches a full terminal chat built with React + Ink: streaming output, tool call visualization, markdown rendering, syntax highlighting, 38 slash commands, 12 themes, and model-aware cost tracking.
Telegram and WhatsApp integration for remote AI interaction. Voice messages are auto-transcribed. Photos are vision-analyzed.
6 event-driven automation types: cron, webhook, file watch, HTTP polling, email (IMAP), and one-time datetime. Events fire AI agent actions with full tool access.
Reusable AI capabilities packaged as YAML frontmatter + Markdown. Progressive loading: metadata at boot, full body on demand.
Third-party commands with sandboxed execution, trust model, and integrity verification.
Layer 4: Relay apps/relay/ (Bun) + apps/relay-worker/ (Cloudflare Worker)
Layer 3: Daemon src/daemon/kernel.ts — 16-step boot, agent, API, services, storage
Layer 2: CLI src/cli/dispatcher.ts — 51 commands, Ink-based interactive REPL
Layer 1: Shared src/shared/ — config, types, output, escape, protocol
Platform src/platform/ — OS abstraction (darwin/linux/win32)
| Entry Point | Description |
|---|---|
jeriko |
Interactive chat REPL |
jeriko <cmd> |
CLI command |
jeriko serve |
Daemon with HTTP API |
TypeScript (strict mode), Bun runtime, compiled to standalone ~66MB binary. React + Ink for terminal UI. SQLite via bun:sqlite. Hono HTTP server.
See docs/ARCHITECTURE.md for the full system design.
| Feature | macOS | Linux | Windows |
|---|---|---|---|
| Core commands, daemon, connectors, chat REPL | Full | Full | Full |
| Document handling (PDF, Excel, Word, images) | Full | Full | Full |
| AI features (all drivers, image gen, STT/TTS) | Full | Full | Partial |
| Native apps (Notes, Reminders, Calendar, etc.) | Full | - | - |
| Desktop screenshot, camera, clipboard | Full | Partial | - |
Build targets: macOS (arm64, x64), Linux (arm64, x64, musl), Windows (x64, arm64)
git clone https://github.com/etheonai/jeriko.git
cd jeriko
bun install
bun run build # standalone binary → ./jeriko
bun run build:all # all platforms → dist/
bun run typecheck # tsc --noEmit
bun test # all testsbun run test:smoke # fast gates (<100ms)
bun run test:unit # 12 parallel subsystem suites
bun run test:integration # relay, commands, connectors
bun run test:e2e # end-to-endCI runs a 5-stage progressive gate: typecheck + smoke, unit (12 parallel jobs), integration, e2e, build verification.
Config lives in ~/.config/jeriko/config.json. Credentials in ~/.config/jeriko/.env (permissions 0600).
jeriko init # interactive setup wizard
jeriko init --ai claude --yes # non-interactiveLoading order: defaults -> user config -> project config -> environment variables.
See docs/INSTALL.md for detailed configuration options.
Defense-in-depth architecture, audited March 2026. Key mechanisms:
- All data stored locally — never transmitted to Etheon
- Single execution gateway with lease/sandbox/audit pipeline
- Sensitive env vars filtered from all subprocesses (3 synchronized lists)
- Shell and AppleScript escaping at 55+ call sites
- Timing-safe HMAC-SHA256 auth at 15+ call sites
- Plugin sandbox with env isolation and integrity verification
- Webhook signature verification (fail-closed)
See docs/SECURITY.md and docs/SECURITY-AUDIT-2026-03-06.md.
| Document | Description |
|---|---|
| AGENT.md | System prompt for AI agents |
| Commands | All 51 commands with flags and examples |
| Architecture | System design, 4-layer model, data flow |
| API Reference | Daemon HTTP API |
| Security | Security model and audit |
| Triggers | 6 trigger types and management |
| Plugins | Plugin system, trust, env isolation |
| Plugin Spec | Plugin manifest and authoring |
| Database | SQLite schema reference |
| Multi-Machine | Distributed orchestration |
| Contributing | Development setup and coding standards |
| Install | Detailed installation guide |
git clone https://github.com/etheonai/jeriko.git
cd jeriko && bun install
bun run dev # watch mode
bun run typecheck # type check
bun test # run testsSee docs/CONTRIBUTING.md for coding standards, naming conventions, and how to add commands.
MIT — see LICENSE for details.
Built by Etheon