PicoClaw
github.com/sipeed/picoclaw- Category
- AI Agents
- Rank
- No. 186Tools index
- Pricing
- Open Source
- Type
- APP
- Use case
- Workflow Automation · Coding
- Interfaces
- CLI · Web · Mobile · MCP · API
- Builder
- @sipeed
- GitHub
- 30.0k stars
- Latest release
- nightly
- Date
About
Ultra-lightweight AI assistant written in Go that runs on $10 hardware with less than 10MB RAM. Designed for deployment on minimal embedded devices while providing full-stack AI capabilities including web search, code development, and task automation.
What it does
PicoClaw connects language models to chat services, a browser console, scheduled jobs, web search, local commands, and external MCP tools. Users can run one-off prompts, interactive sessions, or a persistent gateway that receives messages from supported chat platforms.
Why it's ranked here
The breadth is unusually practical for constrained deployments: many chat channels, configurable model providers, scheduling, health checks, Docker, and MCP support share one Go project. However, the maintainers explicitly warn that development is rapid, security issues may remain, and production deployment should wait until the project reaches its stated stability milestone.
What's good
Installation covers precompiled binaries, source builds, Docker Compose, and a browser launcher. Rule-based model routing can send simpler requests to lighter models. Scheduled work supports agent turns, direct delivery, and shell commands. Remote command scheduling has channel allowlists and confirmation gates. Health, readiness, and reload endpoints support operational monitoring.
Tradeoffs
Recent builds may consume 10 to 20MB despite the lower core-memory claim. The gateway does not provide a generic REST chat endpoint. Running useful workflows requires model credentials, while some search providers need separate keys. The browser console should not face untrusted networks. Native WhatsApp support requires a separate build option and produces a larger binary.
How to use it well
Use PicoClaw for a self-hosted assistant on modest Linux hardware, an Android phone, or a desktop where chat integrations and recurring automation matter. Start with the browser launcher or Docker, configure one model provider, then enable only trusted channels and tools. Do not treat it as a production-hardened public service or a general-purpose REST chat backend.
Technical notes+
README.md documents precompiled downloads, Go 1.25+ source builds, the desktop WebUI launcher, Android support, MCP, vision, and model routing. go.mod declares the Go module and dependencies for Cobra, MCP, SQLite, model providers, and numerous chat SDKs. Makefile builds the core binary from cmd/picoclaw, builds a separate launcher, cross-compiles Linux ARM, ARM64, MIPS, RISC-V, LoongArch, macOS, Windows, and Android targets, and applies a MIPS ELF metadata patch. pkg/health/server.go implements /health, /ready, and POST-only /reload endpoints; reload can require a bearer token checked with constant-time comparison. docs/guides/docker.md states that gateway mode exposes webhook and health handlers, not generic REST chat routes. docs/reference/cron.md details persisted scheduled jobs, execution modes, remote-channel scoping, command confirmation, and allowlists. docs/guides/chat-apps.md lists the supported messaging channels and their transport requirements.
Observed
- License
- MIT
- Primary language
- Go
- Install surface
- Precompiled binaries, source builds with Make, and Docker Compose
- Interfaces
- CLI, browser launcher, chat gateway, and MCP integration
- Platforms
- Linux, macOS, Windows, Android, x86, ARM, MIPS, RISC-V, and LoongArch build targets
- Operations
- HTTP health, readiness, and authenticated reload endpoints
- Scheduling
- Persistent recurring and one-time jobs with agent, delivery, and command execution modes
Read from README.md, go.mod, Makefile, pkg/health/server.go, docs/README.md, docs/guides/README.md, docs/guides/docker.md, docs/reference/cron.md, docs/guides/chat-apps.md, docs/guides/docker.fr.md, docs/guides/docker.ja.md, docs/guides/docker.ms.md, docs/guides/docker.vi.md.
What it can do
Perform web search queries
Search terms or questions → Web search results and relevant information
Generate code from natural language prompts
Natural language description of desired functionality → Source code in various programming languages
Automate tasks based on user instructions
Task descriptions and parameters → Executed automated workflows and results
Provide AI assistance with minimal resource consumption
User queries and commands → AI-generated responses and solutions
Deploy AI capabilities on embedded devices
Deployment configuration and target hardware specs → Running AI assistant on low-resource hardware
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.