AgenticSeek: Private, Local Manus Alternative.

English | 中文 | 繁體中文 | Français | 日本語 | Português (Brasil)

A 100% local alternative to Manus AI, this voice-enabled AI assistant autonomously browses the web, writes code, and plans tasks while keeping all data on your device. Tailored for local reasoning models, it runs entirely on your hardware, ensuring complete privacy and zero cloud dependency.

Why AgenticSeek ?

• 🔒 Fully Local & Private - Everything runs on your machine — no cloud, no data sharing. Your files, conversations, and searches stay private.

• 🌐 Smart Web Browsing - AgenticSeek can browse the internet by itself — search, read, extract info, fill web form — all hands-free.

• 💻 Autonomous Coding Assistant - Need code? It can write, debug, and run programs in Python, C, Go, Java, and more — all without supervision.

• 🧠 Smart Agent Selection - You ask, it figures out the best agent for the job automatically. Like having a team of experts ready to help.

• 📋 Plans & Executes Complex Tasks - From trip planning to complex projects — it can split big tasks into steps and get things done using multiple AI agents.

• 🎙️ Voice-Enabled - Clean, fast, futuristic voice and speech to text allowing you to talk to it like it's your personal AI from a sci-fi movie. (In progress)

Demo

Can you search for the agenticSeek project, learn what skills are required, then open the CV_candidates.zip and then tell me which match best the project

agentic_seek_demo.mov

Disclaimer: This demo, including all the files that appear (e.g: CV_candidates.zip), are entirely fictional. We are not a corporation, we seek open-source contributors not candidates.

🛠⚠️️ Active Work in Progress

🙏 This project started as a side-project and has zero roadmap and zero funding. It's grown way beyond what I expected by ending in GitHub Trending. Contributions, feedback, and patience are deeply appreciated.

Prerequisites

Make sure you have chrome driver, docker and python3.10 installed.

For issues related to chrome driver, see the Chromedriver section.

1. Clone the repository and setup

git clone https://github.com/Fosowl/agenticSeek.git
cd agenticSeek
mv .env.example .env

2. Change the .env file content

SEARXNG_BASE_URL="http://127.0.0.1:8080"
REDIS_BASE_URL="redis://redis:6379/0"
WORK_DIR="/Users/mlg/Documents/workspace_for_ai"
OLLAMA_PORT="11434"
LM_STUDIO_PORT="1234"
CUSTOM_ADDITIONAL_LLM_PORT="11435"
OPENAI_API_KEY='optional'
DEEPSEEK_API_KEY='optional'
OPENROUTER_API_KEY='optional'
TOGETHER_API_KEY='optional'
GOOGLE_API_KEY='optional'
ANTHROPIC_API_KEY='optional'

API Key are totally optional for user who choose to run LLM locally. Which is the primary purpose of this project. Leave empty if you have sufficient hardware

The following environment variables configure your application's connections and API keys.

Update the .env file with your own values as needed:

• SEARXNG_BASE_URL: Leave unchanged
• REDIS_BASE_URL: Leave unchanged
• WORK_DIR: Path to your working directory on your local machine. AgenticSeek will be able to read and interact with these files.
• OLLAMA_PORT: Port number for the Ollama service.
• LM_STUDIO_PORT: Port number for the LM Studio service.
• CUSTOM_ADDITIONAL_LLM_PORT: Port for any additional custom LLM service.

All API key environment variables below are optional. You only need to provide them if you plan to use external APIs instead of running LLMs locally.

3. Start Docker

Make sure Docker is installed and running on your system. You can start Docker using the following commands:

• On Linux/macOS:
  Open a terminal and run:

  sudo systemctl start docker

  Or launch Docker Desktop from your applications menu if installed.

• On Windows:
  Start Docker Desktop from the Start menu.

You can verify Docker is running by executing:

docker info

If you see information about your Docker installation, it is running correctly.

---

Setup for running LLM locally on your machine

Hardware Requirements:

To run LLMs locally, you'll need sufficient hardware. At a minimum, a GPU capable of running Qwen/Deepseek 14B is required. See the FAQ for detailed model/performance recommendations.

Setup your local provider

Start your local provider, for example with ollama:

ollama serve

See below for a list of local supported provider.

Update the config.ini

Change the config.ini file to set the provider_name to a supported provider and provider_model to a LLM supported by your provider. We recommend reasoning model such as Qwen or Deepseek.

See the FAQ at the end of the README for required hardware.

[MAIN]
is_local = True # Whenever you are running locally or with remote provider.
provider_name = ollama # or lm-studio, openai, etc..
provider_model = deepseek-r1:14b # choose a model that fit your hardware
provider_server_address = 127.0.0.1:11434
agent_name = Jarvis # name of your AI
recover_last_session = True # whenever to recover the previous session
save_session = True # whenever to remember the current session
speak = False # text to speech
listen = False # Speech to text, only for CLI, experimental
jarvis_personality = False # Whenever to use a more "Jarvis" like personality (experimental)
languages = en zh # The list of languages, Text to speech will default to the first language on the list
[BROWSER]
headless_browser = True # leave unchanged unless using CLI on host.
stealth_mode = True # Use undetected selenium to reduce browser detection

Warning:

• The config.ini file format does not support comments. Do not copy and paste the example configuration directly, as comments will cause errors. Instead, manually modify the config.ini file with your desired settings, excluding any comments.

• Do NOT set provider_name to openai if using LM-studio for running LLMs. Set it to lm-studio.

• Some provider (eg: lm-studio) require you to have http:// in front of the IP. For example http://127.0.0.1:1234

List of local providers

Provider	Local?	Description
ollama	Yes	Run LLMs locally with ease using ollama as a LLM provider
lm-studio	Yes	Run LLM locally with LM studio (set provider_name to lm-studio)
openai	Yes	Use openai compatible API (eg: llama.cpp server)

Next step: Start services and run AgenticSeek

See the Known issues section if you are having issues

See the Run with an API section if your hardware can't run deepseek locally

See the Config section for detailled config file explanation.

---

Setup to run with an API

Running with an API is optional, see above to run locally.

Set the desired provider in the config.ini. See below for a list of API providers.

[MAIN]
is_local = False
provider_name = google
provider_model = gemini-2.0-flash
provider_server_address = 127.0.0.1:5000 # doesn't matter

Warning: Make sure there is not trailing space in the config.

Export your API key: export <<PROVIDER>>_API_KEY="xxx"

Example: export TOGETHER_API_KEY="xxxxx"

List of API providers

Provider	Local?	Description
openai	Depends	Use ChatGPT API
deepseek	No	Deepseek API (non-private)
huggingface	No	Hugging-Face API (non-private)
togetherAI	No	Use together AI API (non-private)
google	No	Use google gemini API (non-private)

Please note that coding/bash might fail with gemini, it seems to ignore our prompt for format to respect, which are optimized for deepseek r1. Model such are gpt-4o seem to perform poorly with our prompt as well.

Next step: Start services and run AgenticSeek

See the Known issues section if you are having issues

See the Config section for detailled config file explanation.

---

Start services and Run

Start required services. This will start all services from the docker-compose.yml, including: - searxng - redis (required by searxng) - frontend - backend (if using full)

./start_services.sh full # MacOS
start ./start_services.cmd full # Window

Warning: This step will download and load all Docker images, which may take up to 30 minutes. After starting the services, please wait until the backend service is fully running (you should see backend: in the log) before sending any messages. The backend services may take longer to start than others.

Go to http://localhost:3000/ and you should see the web interface.

Optional: Run with the CLI interface:

To run with CLI interface you would have to install package on host:

./install.sh
./install.bat # windows

Start services:

./start_services.sh # MacOS
start ./start_services.cmd # Window

Then run : python3 cli.py

---

Usage

Make sure the services are up and running with ./start_services.sh full and go to localhost:3000 for web interface.

You can also use speech to text by setting listen = True in the config. Only for CLI mode.

To exit, simply say/type goodbye.

Here are some example usage:

Make a snake game in python!

Search the web for top cafes in Rennes, France, and save a list of three with their addresses in rennes_cafes.txt.

Write a Go program to calculate the factorial of a number, save it as factorial.go in your workspace

Search my summer_pictures folder for all JPG files, rename them with today’s date, and save a list of renamed files in photos_list.txt

Search online for popular sci-fi movies from 2024 and pick three to watch tonight. Save the list in movie_night.txt.

Search the web for the latest AI news articles from 2025, select three, and write a Python script to scrape their titles and summaries. Save the script as news_scraper.py and the summaries in ai_news.txt in /home/projects

Friday, search the web for a free stock price API, register with [email protected] then write a Python script to fetch using the API daily prices for Tesla, and save the results in stock_prices.csv

Note that form filling capabilities are still experimental and might fail.

After you type your query, AgenticSeek will allocate the best agent for the task.

Because this is an early prototype, the agent routing system might not always allocate the right agent based on your query.

Therefore, you should be very explicit in what you want and how the AI might proceed for example if you want it to conduct a web search, do not say:

Do you know some good countries for solo-travel?

Instead, ask:

Do a web search and find out which are the best country for solo-travel

---

Setup to run the LLM on your own server

If you have a powerful computer or a server that you can use, but you want to use it from your laptop you have the options to run the LLM on a remote server using our custom llm server.

On your "server" that will run the AI model, get the ip address

ip a | grep "inet " | grep -v 127.0.0.1 | awk '{print $2}' | cut -d/ -f1 # local ip
curl https://ipinfo.io/ip # public ip

Note: For Windows or macOS, use ipconfig or ifconfig respectively to find the IP address.

Clone the repository and enter the server/folder.

git clone --depth 1 https://github.com/Fosowl/agenticSeek.git
cd agenticSeek/llm_server/

Install server specific requirements:

pip3 install -r requirements.txt

Run the server script.

python3 app.py --provider ollama --port 3333

You have the choice between using ollama and llamacpp as a LLM service.

Now on your personal computer:

Change the config.ini file to set the provider_name to server and provider_model to deepseek-r1:xxb. Set the provider_server_address to the ip address of the machine that will run the model.

[MAIN]
is_local = False
provider_name = server
provider_model = deepseek-r1:70b
provider_server_address = x.x.x.x:3333

Next step: Start services and run AgenticSeek

---

Speech to Text

Warning: speech to text only work in CLI mode at the moment.

Please note that currently speech to text only work in english.

The speech-to-text functionality is disabled by default. To enable it, set the listen option to True in the config.ini file:

listen = True

When enabled, the speech-to-text feature listens for a trigger keyword, which is the agent's name, before it begins processing your input. You can customize the agent's name by updating the agent_name value in the config.ini file:

agent_name = Friday

For optimal recognition, we recommend using a common English name like "John" or "Emma" as the agent name

Once you see the transcript start to appear, say the agent's name aloud to wake it up (e.g., "Friday").

Speak your query clearly.

End your request with a confirmation phrase to signal the system to proceed. Examples of confirmation phrases include:

"do it", "go ahead", "execute", "run", "start", "thanks", "would ya", "please", "okay?", "proceed", "continue", "go on", "do that", "go it", "do you understand?"

Config

Example config:

[MAIN]
is_local = True
provider_name = ollama
provider_model = deepseek-r1:32b
provider_server_address = 127.0.0.1:11434
agent_name = Friday
recover_last_session = False
save_session = False
speak = False
listen = False
jarvis_personality = False
languages = en zh
[BROWSER]
headless_browser = False
stealth_mode = False

Explanation:

• is_local -> Runs the agent locally (True) or on a remote server (False).

• provider_name -> The provider to use (one of: ollama, server, lm-studio, deepseek-api)

• provider_model -> The model used, e.g., deepseek-r1:32b.

• provider_server_address -> Server address, e.g., 127.0.0.1:11434 for local. Set to anything for non-local API.

• agent_name -> Name of the agent, e.g., Friday. Used as a trigger word for TTS.

• recover_last_session -> Restarts from last session (True) or not (False).

• save_session -> Saves session data (True) or not (False).

• speak -> Enables voice output (True) or not (False).

• listen -> listen to voice input (True) or not (False).

• jarvis_personality -> Uses a JARVIS-like personality (True) or not (False). This simply change the prompt file.

• languages -> The list of supported language, needed for the llm router to work properly, avoid putting too many or too similar languages.

• headless_browser -> Runs browser without a visible window (True) or not (False).

• stealth_mode -> Make bot detector time harder. Only downside is you have to manually install the anticaptcha extension.

• languages -> List of supported languages. Required for agent routing system. The longer the languages list the more model will be downloaded.

Providers

The table below show the available providers:

Provider	Local?	Description
ollama	Yes	Run LLMs locally with ease using ollama as a LLM provider
server	Yes	Host the model on another machine, run your local machine
lm-studio	Yes	Run LLM locally with LM studio (lm-studio)
openai	Depends	Use ChatGPT API (non-private) or openai compatible API
deepseek-api	No	Deepseek API (non-private)
huggingface	No	Hugging-Face API (non-private)
togetherAI	No	Use together AI API (non-private)
google	No	Use google gemini API (non-private)

To select a provider change the config.ini:

is_local = True
provider_name = ollama
provider_model = deepseek-r1:32b
provider_server_address = 127.0.0.1:5000

is_local: should be True for any locally running LLM, otherwise False.

provider_name: Select the provider to use by it's name, see the provider list above.

provider_model: Set the model to use by the agent.

provider_server_address: can be set to anything if you are not using the server provider.

Known issues

Chromedriver Issues

Known error #1: chromedriver mismatch

Exception: Failed to initialize browser: Message: session not created: This version of ChromeDriver only supports Chrome version 113 Current browser version is 134.0.6998.89 with binary path

This happen if there is a mismatch between your browser and chromedriver version.

You need to navigate to download the latest version:

https://developer.chrome.com/docs/chromedriver/downloads

If you're using Chrome version 115 or newer go to:

https://googlechromelabs.github.io/chrome-for-testing/

And download the chromedriver version matching your OS.

If this section is incomplete please raise an issue.

connection adapters Issues

Exception: Provider lm-studio failed: HTTP request failed: No connection adapters were found for '127.0.0.1:11434/v1/chat/completions'

Make sure you have http:// in front of the provider IP address :

provider_server_address = http://127.0.0.1:11434

SearxNG base URL must be provided

raise ValueError("SearxNG base URL must be provided either as an argument or via the SEARXNG_BASE_URL environment variable.")
ValueError: SearxNG base URL must be provided either as an argument or via the SEARXNG_BASE_URL environment variable.

Maybe you didn't move .env.example as .env ? You can also export SEARXNG_BASE_URL:

export SEARXNG_BASE_URL="http://127.0.0.1:8080"

FAQ

Q: What hardware do I need?

Model Size	GPU	Comment
7B	8GB Vram	⚠️ Not recommended. Performance is poor, frequent hallucinations, and planner agents will likely fail.
14B	12 GB VRAM (e.g. RTX 3060)	✅ Usable for simple tasks. May struggle with web browsing and planning tasks.
32B	24+ GB VRAM (e.g. RTX 4090)	🚀 Success with most tasks, might still struggle with task planning
70B+	48+ GB Vram (eg. mac studio)	💪 Excellent. Recommended for advanced use cases.

Q: Why Deepseek R1 over other models?

Deepseek R1 excels at reasoning and tool use for its size. We think it’s a solid fit for our needs other models work fine, but Deepseek is our primary pick.

Q: I get an error running cli.py. What do I do?

Ensure local is running (ollama serve), your config.ini matches your provider, and dependencies are installed. If none work feel free to raise an issue.

Q: Can it really run 100% locally?

Yes with Ollama, lm-studio or server providers, all speech to text, LLM and text to speech model run locally. Non-local options (OpenAI or others API) are optional.

Q: Why should I use AgenticSeek when I have Manus?

This started as Side-Project we did out of interest about AI agents. What’s special about it is that we want to use local model and avoid APIs. We draw inspiration from Jarvis and Friday (Iron man movies) to make it "cool" but for functionality we take more inspiration from Manus, because that's what people want in the first place: a local manus alternative. Unlike Manus, AgenticSeek prioritizes independence from external systems, giving you more control, privacy and avoid api cost.

Contribute

We’re looking for developers to improve AgenticSeek! Check out open issues or discussion.

Contribution guide

Maintainers:

Fosowl | Paris Time

antoineVIVIES | Taipei Time

steveh8758 | Taipei Time

Special Thanks:

tcsenpai and plitc For helping with backend dockerization