You already run the thing. SearchByAI is where agents and people find it — MCP servers, HTTP APIs and CLI tools, hosted by whoever hosts them. No datacenters. We index your node; we never sit in front of it, so your traffic goes straight to you and never touches our servers.
Every pin is an independently operated CLI, MCP, or API endpoint for AI agent use. Click here to add yours. Drag to rotate.
No active nodes yet — be the first to list one.
Any machine offering something another agent or person can use. Three kinds, described in one manifest.
Tools an agent calls directly. The registry records the transport, URL, and tool names, so a client can be configured from the listing.
Endpoints anything can request. Methods, paths, and schemas are listed, so a caller knows the shape before making a request.
Commands people install and run locally. The install command and usage are listed, so nothing has to be guessed.
The registry stores descriptions. It does not sit between you and the node.
Search returns the node's own endpoint. Your call goes there, not through us — so we never see your payloads and cannot become the bottleneck.
Every capability declares what it can touch — files, network, shell, payments. You see that before you connect, not after.
Nodes serving a physical area are searchable by distance. Location is stored coarsely — about a kilometre, never an address.
This registry lists endpoints you already run. If you have a GPU and nothing on it, install ours first — Ollama, Open WebUI, n8n, LightRAG, crawl4ai and a Cloudflare tunnel, on 8GB of VRAM or more.
Any agent with shell access can do the install. It runs what it can and stops to walk you through the five credentials only a human can issue.
Read https://github.com/aiwtfgpt/searchbyai-public/blob/main/stack/INSTALL.md
and install this stack on my machine.mkdir -p ~/ai-stack && cd ~/ai-stack
curl -fsSL https://raw.githubusercontent.com/aiwtfgpt/searchbyai-public/main/stack/docker-compose.yml \
-o docker-compose.yml
curl -fsSL https://raw.githubusercontent.com/aiwtfgpt/searchbyai-public/main/stack/scripts/install-stack.sh \
-o install-stack.sh
bash install-stack.shThen INSTALL.md from step 2 for the tunnel, Google, n8n and Claude credentials.
Everything is public — read it before you run it: github.com/aiwtfgpt/searchbyai-public · what gets installed
One command for a person. One document for an agent. Both produce the same manifest.
Run this on the machine you want to list. It asks a few questions, registers the node, and sets up heartbeats.
bash <(curl -fsSL https://raw.githubusercontent.com/aiwtfgpt/searchbyai-public/main/registry/connect.sh)
Requires Linux or macOS, plus curl, jq,
openssl, and cron. Not supported on native Windows —
use WSL.
Read the onboarding document. It walks through inspecting the host, writing the manifest, and registering.
https://github.com/aiwtfgpt/searchbyai-public/blob/main/registry/CONNECT.mdOne JSON endpoint. Text, capability kind, and distance.
# find OCR nodes within 100km of Denver
curl 'https://searchbyai.com/api/discover?q=ocr&near=39.74,-104.99&radius_km=100'
# every MCP server in the registry
curl 'https://searchbyai.com/api/discover?kind=mcp'
# full detail for one node — same URL a human would open
curl -H 'Accept: application/json' https://searchbyai.com/n/some-nodeNode pages serve HTML to browsers and JSON to agents from the same address, so one link works for both.
Occasional updates on the registry. No noise.