Index / TinyAGI (fka TinyClaw) · updated Oct 24, 2026
TinyAGI (fka TinyClaw)
TinyAGI/tinyclaw · healthy · rank 21 of 50 by stars
A multi-agent, multi-team orchestrator aimed at running a 'one person company' with isolated agent workspaces and a slick web office UI. Actively developed with rapid feature churn around its TinyOffice dashboard, task system, and channel integrations.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 55 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 32 files examined
1 provider · custom endpoint
Pinned models
gpt-5.3-codex released 2026-02-24 — 8 months old (from the public model catalogue)
Pin last edited 7 months ago
Evidence
Decision
Why choose TinyAGI (fka TinyClaw) over OpenClaw?
Why choose this
- Native multi-team orchestration with agent hand-offs and fan-out
- Rich TinyOffice web UI with task/project management and live visualizer
- Built-in multi-channel support (Discord, WhatsApp, Telegram)
Tradeoffs
- Experimental stability versus OpenClaw's more mature core
- Smaller skills/plugin ecosystem than OpenClaw's registry
- Heavier Node.js stack and more moving parts to operate
Best fit
- Solo founders wanting multiple collaborating agent teams
- Users who want a polished web dashboard for agent management
- Teams needing Discord/WhatsApp/Telegram agent channels
Avoid if
- You need a stable, production-ready platform (marked experimental)
- You want a lightweight single-agent assistant
- You require strong sandboxing guarantees out of the box
Evidence is solid for features and activity (detailed README, frequent recent commits, v0.0.20 release), but community sentiment data is thin with only one tangential Reddit mention and mostly aggregator-digest web results.
AI layer reviewed Oct 5, 2026 · how this is written
Star activity
3,626 stars today
Overview
TinyAGI (formerly TinyClaw) is a TypeScript-based multi-agent orchestrator designed around the 'One Person Company' concept: you run multiple teams of AI agents with specialized roles that collaborate via chain execution and fan-out, each in isolated workspaces. It supports Anthropic Claude and OpenAI Codex providers (plus any OpenAI/Anthropic-compatible endpoint), persists sessions in SQLite with an atomic queue, retry logic, and dead-letter management, and runs 24/7 as a daemon or Docker container.
Its standout differentiator versus OpenClaw is the TinyOffice web portal — a browser dashboard with a pixel-office agent visualizer, Linear-style task/project management with comments, team chat rooms, and a control plane for services, providers, pairing, and logs. Recent commits show heavy investment in this UI layer, plus a plugin system for message hooks and event listeners.
Compared to OpenClaw, TinyAGI trades ecosystem maturity for orchestration breadth: it is explicitly experimental, its plugin ecosystem is emerging rather than established, and some operational choices (e.g., running as root in Docker, delegating isolation to gVisor) raise the security bar for self-hosters. It fits users who want a managed multi-agent 'company' experience more than a single local assistant.