Index / SmallClaw · updated Sep 3, 2026

SmallClaw

XposeMarket/SmallClaw · healthy · rank 34 of 43 by stars

A local-first, chat-first AI agent framework tuned for small local models, with a single-pass architecture that skips multi-agent planning overhead. It positions itself as a free, no-API-cost OpenClaw alternative with optional hybrid cloud support.

#local-first#ai-agent#ollama#small-models#openclaw-alternative
Compare vs OpenClaw
GitHub ↗

Facts

Repository

Contributors1
Open issues4
Last commitMar 5, 2026

Runtime

LanguageTypeScript
Memory80 MB
Boot time180 ms
Deploymentdesktop · self-hosted
Setup difficultyMedium
Plugin ecosystemEmerging

Posture

Local-firstYes
Cloud dependencyOptional
Privacy postureStrong

Community

Sentiment35% positive
Reddit mentions15
Web results8

Security breakdown

Composite 62 / 100 · how these are scored

Sandboxing5 / 10

higher is safer

API security6 / 10

higher is safer

Network isolation6 / 10

higher is safer

Telemetry safety7 / 10

higher is safer

Shell access risk7 / 10

higher is riskier

Model access

Read from the repository, not written by a model · 17 files examined

1 provider · runs locally

Direct
OpenAI
Local
Ollama

Pinned models

gpt-4oOpenAI

gpt-4o released 2024-05-13 — 2 years old (from the public model catalogue)

Pin last edited 6 months ago

Several defaults are pinned; the project picks one per provider rather than shipping a single default.

Evidence

Decision

Why choose SmallClaw over OpenClaw?

Why choose this

  • Single-pass chat handler reduces latency on small models
  • Explicitly optimized for small local models
  • No API costs with fully local provider support

Tradeoffs

  • Much smaller community and ecosystem
  • License unclear; copyright notice suggests proprietary terms
  • Fewer channels and integrations than OpenClaw

Best fit

  • Users wanting a free local agent with no API costs
  • Running small local models via Ollama or llama.cpp
  • Privacy-conscious users who want data on-device

Avoid if

  • You need a proven, audited security model
  • You want a large established community and plugin registry
  • You need multi-user or team deployments
Good Confidence55%

README and commit history give a decent picture of architecture and features, but the license is ambiguous and community sentiment data is essentially noise (no genuine Reddit discussion found). Confidence in decision fields is moderate.

AI layer reviewed Aug 24, 2026 · how this is written

Star activity

259 stars today

Overview

SmallClaw is a local-first AI agent framework from Xpose Market, built in TypeScript and designed specifically to work well with small local models. It supports multiple providers — Ollama, llama.cpp, LM Studio, OpenAI API, and OpenAI Codex OAuth — allowing fully local or hybrid cloud setups, and ships with a web UI dashboard for chat-first interaction.

Its core architectural bet is a single-pass chat handler: one LLM call decides whether to respond conversationally or invoke tools, rather than running separate planning, execution, and verification agents. This reduces latency and plays to the strengths of small models that struggle with complex multi-agent coordination. Tooling includes file operations with line-level edits, multi-provider web search (Tavily, Google, Brave, DuckDuckGo), Playwright browser automation, terminal access, persistent session memory, and a drop-in SKILL.md skills system.

Compared to OpenClaw, SmallClaw trades ecosystem breadth for simplicity and local-model efficiency. Recent commits show active development — v1.1 brought a system overhaul, and v1.0.4 added security hardening, Docker improvements, background tasks, sub-agents, and a Telegram file browser. However, the license situation is unclear (the description asserts 'All rights reserved'), and there is little genuine community discussion, so prospective users should verify terms and security posture before adopting it for sensitive workflows.

Nominate a clone

Add a new Claw

Paste a GitHub repository and tell us why it belongs on the tracker.

Opens a prefilled issue on GitHub — every nomination is public. Comfortable with a PR? Adding the repo to projects.json is faster.