AgenticSeek
Open-source local-first AI agent that browses the web, writes code, and runs tasks without sending anything to the cloud.
Pick AgenticSeek if you want a Manus-style autonomous agent that never touches the cloud and you already run local LLMs.
Skip it if you want a turnkey hosted agent, frontier-model reasoning, or enterprise support — there is no SaaS layer here.
AgenticSeek is a fully local, open-source AI assistant that bills itself as a privacy-respecting alternative to Manus, Devin, and the various cloud-hosted agent stacks. It runs entirely on your own machine, orchestrating a small fleet of specialised agents that can browse the web autonomously, write and execute code in Python, C, Go and Java, plan multi-step projects, and respond to voice commands.
What sets it apart is the no-cloud, no-API posture: everything from model inference to file access stays on your hardware, which makes it appealing to developers working with sensitive code, journalists, researchers and anyone who refuses to ship prompts to OpenAI or Anthropic. The project is free and MIT-style open source, distributed via GitHub, so the only real cost is the GPU you point it at. An optional SerpApi key unlocks richer web search and flight lookups, but the core loop works without any third-party keys.
The trade-off is that AgenticSeek is a hobbyist-grade project rather than a polished product: you bring your own local LLM runtime (typically Ollama or a similar llama.cpp wrapper), and quality of output is bounded by whatever open-weight model you can run on your box. There is no managed SaaS, no support contract, and the marketing site is a single GitHub Pages doc.
AgenticSeek is one of the more credible open-source attempts at a privacy-first answer to Manus and Devin. It will not match a GPT-4-class agent on raw capability, but for tinkerers with a decent GPU and a hard no-cloud rule, it is genuinely useful and refreshingly honest about what it is.
— The AI Tool Bible editorial team
Pros
- ✅ 100% local execution — no prompts or files leave your machine
- ✅ Free and open source under a permissive license
- ✅ Multi-agent architecture covering web, code, planning and voice
- ✅ Works with whatever open-weight model your hardware can run
Cons
- ⚠️ Output quality is capped by the local model you can host
- ⚠️ Requires technical setup; not a polished consumer product
- ⚠️ Sparse public docs and no managed/hosted option
- ⚠️ Performance demands a capable GPU for usable latency
Use cases
Explore related
Compare with similar tools
All in Agents →LangGraph
FeaturedStateful, graph-based agent orchestration from LangChain.
CrewAI
FeaturedPython framework for multi-agent orchestration.
Claude Agent SDK
Anthropic's official SDK for building autonomous Claude agents.
Manus
Generalist agent for research, code, and web tasks.
Devin
Cognition Labs' "autonomous software engineer" agent.
AutoGPT
Open-source platform for building autonomous AI agents.