CoexistAI MCP
A local research MCP server: web, Reddit and YouTube search, maps and code exploration for a model in LM Studio or Cursor
- 530 ★ on the repository
- Last commit
Extended access
Why this level
- Reads local folders and files at the given paths, and their contents go to the chosen language model and, in cloud mode, to the provider
- Goes online: downloads web pages and searches through a SearXNG container
- The setup script starts a search container and installs Python dependencies
Install
- Works in
Manual install
git clone https://github.com/SPThole/CoexistAI.git coexistai
cd coexistai
bash quick_setup.shConfigure config/model_config.json first. Docker must be running. On macOS with zsh use zsh quick_setup.sh.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Reads local folders and files at the given paths, and their contents go to the chosen language model and, in cloud mode, to the provider
- Goes online: downloads web pages and searches through a SearXNG container
- The setup script starts a search container and installs Python dependencies
This is third-party code. Review the repository files before installing.
What it does
CoexistAI runs on your machine as a FastAPI server and opens an MCP server alongside it. It gathers several sources into one toolset. Web search summarizes the pages it finds, Reddit search ranks discussions with BM25, and YouTube videos are transcribed and summarized by link or search phrase. Maps build routes and show nearby points of interest. Other tools work with code and files. A GitHub repository or a local git folder can be browsed as a tree and queried with questions. Local folders and files are read together with images, and the document mode of web search supports PDF. Two extra modes turn text into speech and into a podcast. You choose the model: Google Gemini, OpenAI, a local one through Ollama or LM Studio, or any provider with a compatible API. Embeddings can be local. Internet search goes through a SearXNG container that the setup script starts. The same functions are available through HTTP requests with a Swagger page and as a Python library.
Who it is for. For those who want to give a local model in LM Studio or Cursor a set of research tools and keep the service on their own machine.
Good fit when
- A local model in LM Studio needs web search that runs sub-queries, ranks results and summarizes pages itself
- You want to study Reddit discussions by phrase or in a specific subreddit and get a summary
- You want a short retelling of a YouTube video, or to find videos on a topic and question their contents
- You want to ask questions about someone else's GitHub repository or a local code folder without cloning it by hand
- You are planning a trip and want a route map with hotels or cafes nearby
Not a fit when
- You want a ready cloud service with no installation: CoexistAI is deployed on your own machine
- You want a single lightweight search tool: CoexistAI brings Docker, a SearXNG container and a Python environment
- You need to work through the official Reddit API: the project takes data from public Reddit JSON endpoints
- You want an agent that writes the final report itself: CoexistAI gives the model tools and leaves the conclusions to it
Example request
Find Reddit discussions of Gemma 3N from the past month, list users' main complaints and add links to the most detailed threadsMCP
- Transport
- http
- Authentication
- not required
Environment variables
GOOGLE_API_KEYsecret- Google Gemini key. Needed only in google mode; for other providers the key goes into config/model_config.json.
Limitations
- The license is custom: non-commercial research, prototyping and education only, and commercial and production use is prohibited.
- Docker with a running daemon, Python 3 and pip are required. For Windows the author suggests WSL or Git Bash.
- Modes with cloud models use the provider's API, Google Gemini by default. OpenAI APIs are unavailable from Russia. Local models through LM Studio or Ollama do not depend on external keys.
- For MCP the author recommends a model that produces structured output quickly and cleanly. Weaker local models do worse.
- The project takes Reddit data from public endpoints, and the user is responsible for following site rules.
How to disable. Remove the coexistai entry from mcp.json in LM Studio or another client. Then stop the FastAPI server and the SearXNG container, and delete the project folder and virtual environment by hand.
FAQ
How do I connect CoexistAI to LM Studio?
Start the server with quick_setup.sh, then in LM Studio (version 0.3.17 or newer) open Install, Edit mcp.json and add a coexistai server with the address http://127.0.0.1:8000/mcp. The author also advises pasting the contents of system_prompt.py into the system prompt.
Can I use CoexistAI with local models and no cloud keys?
Yes. In config/model_config.json set the type to local with your server's address, for example LM Studio at http://localhost:1234/v1, and take local embeddings through infinity_emb. For LM Studio the author suggests the Qwen3-8B-GGUF and gemma-3-12b models.
What tools does CoexistAI provide?
Web search and page summaries, YouTube search and video transcripts, Reddit search, maps and routes, a tree and questions for a GitHub repository or local git, exploration of local files with images, and text to speech and podcast.
Can I use CoexistAI in a commercial project?
No. The license allows only non-commercial research, prototyping and education, and commercial and production use of the project and its outputs is prohibited.
Where can I see the available requests without MCP?
After the server starts, open Swagger at http://127.0.0.1:8000/docs. It describes, for example, POST /web-search and POST /web-summarize.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server that searches Google and Scholar, reads pages and PDFs and stores findings in a local project knowledge graph
A unified search MCP server: one interface to Tavily, Brave, Kagi, Exa, Linkup, Firecrawl and GitHub search
Yandex and Google search MCP server
web-compass
An MCP web-search and content-extraction server via Yandex and Google on top of SerpAPI, geared toward RU/CIS results
A web search MCP server that gives the agent full page text at once: Stack Overflow answers, GitHub threads, arXiv papers
A metasearch MCP server: the agent searches text, images, news, videos and books across many engines and extracts page content
Official Oxylabs MCP server: page scraping with JS rendering, Google and Amazon results, an AI crawler
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| CoexistAI (this entry) | no | not required | 530 | |
| GPT Researcher | no | 30,357 | an autonomous agent that writes a final report on a question, whereas CoexistAI hands a model a set of tools | |
| SearXNG MCP Server | no | not required | 1,338 | web search through SearXNG only, no Reddit, YouTube, maps or repositories |
| Wikipedia MCP Server | no | not required | 303 | a Wikipedia MCP server |
README in short
The README begins with a list of new features and capabilities, then describes installation through Docker and locally. Most of it gives request examples for each of seven endpoints and Python usage, and the end covers the user's responsibility for following source rules.