Open-source voice dictation app for Mac: hold Fn, speak, and cleaned-up text is typed wherever your cursor is.
Fosowl/agenticSeek
A fully local, voice-enabled AI assistant that browses the web, writes code and plans tasks on your own hardware, as an alternative to Manus AI.
Open-source alternative to:Manus
- Stars
- 27,530
- Forks
- 3,085
- Open issues
- 33
- Last commit
- 4 hours ago
What is it?
AgenticSeek is a self-hosted autonomous agent that takes a natural-language request, picks an appropriate sub-agent, and carries out the work: searching and reading web pages, filling in web forms, writing and running code in languages such as Python, C, Go and Java, and splitting larger goals into steps. It is designed to run on local reasoning models such as DeepSeek-R1 or Magistral served through Ollama or LM Studio, so files, conversations and searches stay on your machine.
The stack runs mainly in Docker: a SearXNG instance for search, Redis, a web frontend and a backend. Agent-generated code runs as an unprivileged user with API keys removed from its environment, and the agent can read and write only the directory you set as WORK_DIR. Cloud providers such as OpenAI, Google Gemini, DeepSeek, TogetherAI and OpenRouter can be used instead of local models by changing config.ini. A CLI mode runs the backend on the host and adds optional speech-to-text and text-to-speech.
The project is GPL-3.0 licensed and is at an early prototype stage, so agent routing and browser form filling can fail on some tasks.
Key features
- 01Autonomous web browsing, searching, and form filling through a browser agent.
- 02Writes, debugs and runs code in Python, C, Go, Java and other languages.
- 03Routes each request to a suitable agent and splits complex tasks into steps.
- 04Works with local models via Ollama, LM Studio or any OpenAI-compatible server.
- 05Optional cloud providers such as OpenAI, Gemini, DeepSeek and OpenRouter.
- 06Runs in Docker with a web interface, plus a CLI mode and optional speech input.
Use cases
- Running a private research assistant that searches the web and saves findings to local files.
- Generating and running scripts, such as a web scraper or a small game, in a controlled work directory.
- Automating repetitive file tasks like renaming photos or compiling lists from a local folder.
- Experimenting with local reasoning models as an agent backend without sending data to cloud APIs.
Quick start
- 1
Clone the repository and copy the environment file
git clone https://github.com/Fosowl/agenticSeek.git && cd agenticSeek && mv .env.example .env - 2
Edit the .env and config.ini files with your provider and paths
- 3
Start Docker and make sure it is running
docker info - 4
Start the services and run AgenticSeek (web interface)
./start_services.sh full
Things to know
- Needs strong local hardware; a GPU capable of running a 14B model is the minimum, or you must pay for an API key.
- Requires Docker, Python 3.10.x and a local LLM provider such as Ollama, so setup takes real effort.
- Voice features are still in progress and listed as experimental.
In collections
Topics
FAQ
- What is agenticSeek?
- A fully local, voice-enabled AI assistant that browses the web, writes code and plans tasks on your own hardware, as an alternative to Manus AI. AgenticSeek is a self-hosted autonomous agent that takes a natural-language request, picks an appropriate sub-agent, and carries out the work: searching and reading web pages, filling in web forms, writing and running code in languages such as Python, C, Go and Java, and splitting larger goals into steps. It is designed to run on local reasoning models such as DeepSeek-R1 or Magistral served through Ollama or LM Studio, so files, conversations and searches stay on your machine.
- Is agenticSeek free to use? What license does it have?
- agenticSeek is open source under the GPL-3.0 license. Check the license terms before using it in commercial products.
- Is agenticSeek still maintained?
- Actively maintained. The last commit to agenticSeek was on Oct 22, 2026.
- What can I use agenticSeek for?
- Running a private research assistant that searches the web and saves findings to local files. Generating and running scripts, such as a web scraper or a small game, in a controlled work directory. Automating repetitive file tasks like renaming photos or compiling lists from a local folder. Experimenting with local reasoning models as an agent backend without sending data to cloud APIs.
- How do I install agenticSeek?
- Clone the repository and copy the environment file: git clone https://github.com/Fosowl/agenticSeek.git && cd agenticSeek && mv .env.example .env Edit the .env and config.ini files with your provider and paths Start Docker and make sure it is running: docker info Start the services and run AgenticSeek (web interface): ./start_services.sh full
- Where can I get agenticSeek?
- The source code and installation instructions are on GitHub: https://github.com/Fosowl/agenticSeek
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A personal AI assistant that runs on your own computer, handling long tasks, remembering context, and talking by voice with any local or cloud model.
Self-hosted AI research assistant that runs deep, multi-step searches across local and cloud LLMs and many search engines, with citations and encrypted storage.
A local-first browser automation engine and Agent Skill that runs Qwen3.5 models on-device to pick page actions from visible elements.
Air-gapped, open-source desktop app: a NotebookLM-style AI agent that researches, transforms and edits your local documents.
A self-hosted Go AI agent that runs 24/7 on your own computer, auto-builds missing tools, and shares them with Claude Code and other agents via MCP.