AI SummaryTinyClaw enables developers to run multiple specialized AI agents simultaneously with isolated workspaces and independent configurations, allowing for task-specific agent teams that can be routed via simple @agent_id syntax. This is valuable for projects requiring diverse AI capabilities (coding, writing, analysis) working in parallel without state interference.
Install
Copy this and paste it into Claude Code, Cursor, or any AI assistant:
I want to set up the "Agents" agent in my project. Please run this command in my terminal: # Add AGENTS.md to your project root curl --retry 3 --retry-delay 2 --retry-all-errors -o AGENTS.md "https://raw.githubusercontent.com/TinyAGI/tinyclaw/main/docs/AGENTS.md" Then explain what the agent does and how to invoke it.
Description
TinyClaw is a team of personal agents that collaborate with each other
Overview
The agent management feature enables you to: • Run multiple agents with different models, providers, and configurations • Route messages to specific agents using @agent_id syntax • Isolate conversations - each agent has its own workspace directory and conversation history • Specialize agents - give each agent a custom system prompt and configuration • Switch providers - mix Anthropic (Claude) and OpenAI (Codex) agents • Customize workspaces - organize agents in your own workspace directory
Agents
TinyClaw supports running multiple AI agents simultaneously, each with its own isolated workspace, configuration, and conversation state. This allows you to have specialized agents for different tasks while maintaining complete isolation.
Architecture
`text ┌─────────────────────────────────────────────────────────────┐ │ Message Channels │ │ (Discord, Telegram, WhatsApp) │ └────────────────────┬────────────────────────────────────────┘ │ │ User sends: "@coder fix the bug" ↓ ┌─────────────────────────────────────────────────────────────┐ │ Queue Processor │ │ • Parses @agent_id routing prefix │ │ • Falls back to default agent if no prefix │ │ • Loads agent configuration from settings.json │ └────────────────────┬────────────────────────────────────────┘ │ ↓ ┌─────────────────────────────────────────────────────────────┐ │ Agent Router │ │ │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ │ │ @coder │ │ @writer │ │ @assistant │ │ │ │ │ │ │ │ (default) │ │ │ │ Provider: │ │ Provider: │ │ Provider: │ │ │ │ anthropic │ │ openai │ │ anthropic │ │ │ │ Model: │ │ Model: │ │ Model: │ │ │ │ sonnet │ │ gpt-5.3-codex│ │ opus │ │ │ │ │ │ │ │ │ │ │ │ Workspace: │ │ Workspace: │ │ Workspace: │ │ │ │ ~/workspace/ │ │ ~/workspace/ │ │ ~/workspace/ │ │ │ │ coder/ │ │ writer/ │ │ assistant/ │ │ │ │ │ │ │ │ │ │ │ │ Config: │ │ Config: │ │ Config: │ │ │ │ .claude/ │ │ .claude/ │ │ .claude/ │ │ │ │ heartbeat.md │ │ heartbeat.md │ │ heartbeat.md │ │ │ │ AGENTS.md │ │ AGENTS.md │ │ AGENTS.md │ │ │ └──────────────┘ └──────────────┘ └──────────────┘ │ │ │ │ Shared: ~/.tinyclaw/ (channels, files, logs, tinyclaw.db) │ └─────────────────────────────────────────────────────────────┘ `
1. Message Routing
When a message arrives, the queue processor parses it for routing: `typescript // User sends: "@coder fix the authentication bug" const routing = parseAgentRouting(rawMessage, agents); // Result: { agentId: "coder", message: "fix the authentication bug" } ` Routing Rules: • Message starts with @agent_id → Routes to that agent • No prefix → Routes to default agent (user-named during setup) • Agent not found → Falls back to default agent • No agents configured → Uses legacy single-agent mode
Discussion
Health Signals
My Fox Den
Community Rating
Sign in to rate this booster