AgenticSeek

AgenticSeek is a local alternative to Manus AI for people who want an assistant to carry out research, coding and file tasks on their own hardware. It can break a request into steps and assign the work to specialized agents.

• Web research: The agent searches the web, reads pages and extracts information that it can save to local files.
• Coding: It writes, debugs and runs programs in languages including Python, C, Go and Java.
• File tasks: It can work with files in a designated workspace, including finding images, renaming them and saving lists of results.
• Task planning: It divides larger requests into smaller jobs and selects agents to carry them out, though its routing can choose the wrong agent.

The free, open-source project uses the GPL-3.0 license. It runs on macOS, Windows and Linux, with a Docker-based web interface or a command-line interface. Local model backends include Ollama, LM Studio and OpenAI-compatible servers such as llama.cpp. The model can also run on your own server.

Local inference doesn't require a cloud API key and keeps model processing on hardware you control. Web research still needs internet access. Optional cloud providers include OpenAI, Google Gemini, DeepSeek, OpenRouter and MiniMax; those connections require API keys and send model requests to the chosen service.

Local use requires a GPU capable of running a reasoning model such as Magistral, Qwen or DeepSeek at the 14B scale. The prompts favor DeepSeek-style models; OpenAI models can struggle with complex browsing and planning, while Gemini can have formatting issues in coding tasks.