TinyClaw
github.com/jlia0/tinyclaw- Category
- AI Agents
- Rank
- No. 442Tools index
- Pricing
- Open Source
- Platform
- web · cli · desktop
- Type
- APP
- Builder
- @jlia0
- GitHub
- 3.6k stars
- Latest release
- v0.0.20
- Date
About
A multi-agent AI platform that enables teams of specialized AI agents to collaborate with each other across multiple channels like Discord, WhatsApp, and Telegram. Features a web dashboard, persistent chat rooms, and 24/7 operation capabilities.
What it does
TinyClaw runs a local message-processing service that assigns requests to configured agents and teams. Each agent gets its own workspace, instructions, model provider, memory, and conversation context. A SQLite-backed queue coordinates work, while browser, terminal, chat-channel, and HTTP interfaces expose control and activity.
Why it's ranked here
The project combines unusually broad operational tooling with a clear experimental warning. Agent routing, team leadership, persistent state, retries, dead-letter handling, scheduling, live events, and graceful shutdown form a credible working system. However, required provider CLIs, local process management, and channel setup make it better suited to hands-on operators than casual users.
What's good
Agent isolation is concrete: each agent receives a separate working directory, custom instructions, synchronized skills, and indexed memory. Queue processing serializes work per agent while allowing different agents to run concurrently. Atomic settings writes, stale-job recovery, maintenance pruning, provider switching without restart, and persistent team chat add practical resilience.
Tradeoffs
The README labels the project experimental. Anthropic and OpenAI operation depends on their respective command-line tools, while custom providers still use one of those harnesses. Windows support is limited to WSL2. Channel integrations require separate bot credentials or device pairing, and the quick installer pipes a remote shell script directly into Bash.
How to use it well
Use it when a technical team wants locally operated agents with distinct roles, durable context, inspectable queues, and chat-based routing. Start with two narrowly defined agents, assign a clear team leader, then monitor logs and failed jobs before enabling unattended operation. It does not replace provider access, channel administration, or general infrastructure monitoring.
Technical notes+
package.json defines a TypeScript npm workspace spanning packages and exposes the command-line package through its bin configuration. packages/main/src/index.ts initializes the SQLite queue, HTTP server, channels, heartbeat, scheduler, plugin loading, per-agent promise chains, stale-message recovery, pruning, and signal-driven shutdown. packages/core/src/agent.ts assembles cached prompts from built-in instructions, teammate metadata, memory indexes, workspace instructions, and configured prompts. packages/cli/src/shared.ts performs atomic settings replacement through a temporary file and rename. packages/server/src/sse.ts forwards internal events to connected SSE clients, while packages/teams/src/index.ts re-exports routing, conversation, and chatroom orchestration.
Observed
- License
- MIT, stated by the README badge
- Primary language
- TypeScript
- Packaging
- npm workspaces monorepo with a global command-line entry
- Install surfaces
- Shell installer, source build, and Docker Compose
- Interfaces
- CLI, browser portal, local HTTP API, SSE event stream, and live terminal viewers
- Platform support
- macOS, Linux, and Windows through WSL2
Read from README.md, package.json, packages/cli/src/logs.ts, packages/cli/src/team.ts, packages/core/src/ids.ts, packages/cli/src/agent.ts, packages/core/src/index.ts, packages/main/src/index.ts, packages/cli/src/daemon.ts, packages/cli/src/shared.ts, packages/cli/src/update.ts, packages/core/src/agent.ts, packages/core/src/types.ts, packages/server/src/sse.ts, packages/teams/src/index.ts.
What it can do
Deploy AI agent teams to messaging platforms
AI agent configurations and platform credentials → Active AI agents operating on Discord, WhatsApp, and Telegram
Create isolated workspaces for different AI agent teams
Team configuration parameters and workspace requirements → Separate working environments with isolated agent operations
Enable AI agents to collaborate within teams
Multi-agent task assignments and collaboration rules → Coordinated responses and shared problem-solving between agents
Maintain persistent chat rooms across channels
Chat history and conversation context → Continuous conversation threads that persist across sessions
Monitor and manage AI agent operations through web dashboard
Agent performance data and system metrics → Real-time oversight and control interface for agent activities
Run AI agent teams continuously without interruption
System resources and operational parameters → 24/7 automated AI assistance across all configured channels
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.