What makes Clawlet ultra-lightweight and efficient?
Clawlet is designed as a single static binary with no runtime dependencies or CGO, and includes bundled SQLite. This architecture allows it to be easily dropped onto any machine and run efficiently without requiring extensive setup or external libraries, making it highly portable and resource-friendly.
Which LLM providers does Clawlet support?
Clawlet supports a variety of LLM providers, including OpenAI, OpenRouter, Anthropic, and Gemini. It also allows integration with local models via Ollama or vLLM, providing flexibility for users to choose their preferred language model backend.
How does Clawlet handle memory and context?
Clawlet features hybrid semantic memory search, enabled by bundled SQLite and sqlite-vec. When memory search is enabled, the agent gains tools to retrieve past context. It indexes Markdown files like MEMORY.md and YYYY-MM-DD.md, injecting them into the user context for more informed interactions.
Can Clawlet be integrated with existing chat applications?
Yes, Clawlet offers integrations with popular chat applications such as Telegram, WhatsApp, Discord, and Slack. These integrations allow users to interact with their personal AI assistant directly through their preferred messaging platforms, often without requiring public webhook endpoints.
What security measures are in place for Clawlet?
Clawlet implements secure defaults, including restricting tools to the workspace directory, binding the gateway to localhost only, and blocking unsafe shell constructs in its exec tool. It also prevents path traversal and sensitive state path access, enhancing overall security.