@noticed/cli
CLI and stdio MCP server for noticed. Exposes ~50 noticed agent capabilities (network search across GitHub + LinkedIn, missions, PRM, memory, workspace, web search, cron, ...) via two meta-tools — search + execute — to any MCP-aware client (Claude Code, Claude Desktop, Cursor, ChatGPT).
README
@noticed/cli
CLI, MCP server, and Claude Code plugin for noticed — search your developer network, trace connections, and find the shortest path to anyone through GitHub and LinkedIn collaboration graphs.
npm install -g @noticed/cli
noticed search "AI engineers"
noticed path @sarahml
Add to your AI coding agent
The MCP server exposes two meta-tools — search and execute — backed by ~50 noticed capabilities: developer-network search and connection paths, mission and goal tracking, a PRM (people-relationship-management) board, a virtual filesystem for agent workspace files, persistent memory, web search, scheduled crons, and more. Same surface the noticed web and Telegram agents use. Chat-only capabilities (in-chat messaging, referral invites, the Cursor Cloud bridge) are filtered server-side.
Upgrading from 0.2.x? The tool surface changed: clients that called
search_network/get_connection_pathdirectly should now callsearch(to discover the capability) followed byexecute { capability: "search_network", args: { query: "…" } }. MCP-aware LLMs handle this discovery automatically.
You have two ways to connect: hosted (no install, Streamable HTTP) or stdio (this package via npx). The hosted server runs your queries against noticed.so so anyone with a noticed account can use it. The stdio server is useful when your MCP client can't speak HTTP, or when you're running a self-hosted noticed instance.
Hosted MCP server (recommended — no install)
https://mcp.noticed.so/api/mcp is a hosted Streamable HTTP endpoint. Two authentication paths are supported; pick whichever your client prefers.
OAuth (claude.ai web, ChatGPT MCP, anything that does OAuth discovery)
The hosted server is a spec-compliant OAuth 2.1 authorization server with Dynamic Client Registration. Clients that follow the MCP authorization profile — including claude.ai's custom connectors and ChatGPT's MCP integration — discover the server, register themselves, and walk the user through a consent screen automatically. No API key needed in the client config.
claude.ai custom connector: open Settings → Connectors → Add custom connector, paste https://mcp.noticed.so/api/mcp as the URL. claude.ai handles the rest. Connected applications are visible (and revocable) at noticed.so/dashboard/oauth-grants.
API key Bearer (Claude Code, Cursor, anything that supports a custom header)
For clients that can set an Authorization: Bearer … header, mint a key at noticed.so/dashboard/api-keys and use it directly — no OAuth flow needed.
Claude Code (one command):
claude mcp add --transport http --scope user noticed https://mcp.noticed.so/api/mcp --header "Authorization: Bearer nk_live_…"
Cursor, Claude Desktop, Zed, VS Code Copilot, Windsurf, Cline — all support URL + header config. Drop the command/args/env block from any of the stdio snippets below and replace with:
{
"mcpServers": {
"noticed": {
"url": "https://mcp.noticed.so/api/mcp",
"headers": { "Authorization": "Bearer nk_live_…" }
}
}
}
Stdio MCP server (this package)
Pick your client below for the stdio install.
Claude Code
claude mcp add --scope project noticed -- npx -y @noticed/cli mcp
--scope project writes to .mcp.json at your repo root so the server is shared with everyone on the team. Drop the flag for a personal-scope install.
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"noticed": {
"command": "npx",
"args": ["-y", "@noticed/cli", "mcp"],
"env": {
"NOTICED_API_KEY": "nk_live_…"
}
}
}
}
Cursor
Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json (project):
{
"mcpServers": {
"noticed": {
"command": "npx",
"args": ["-y", "@noticed/cli", "mcp"],
"env": { "NOTICED_API_KEY": "nk_live_…" }
}
}
}
VS Code (Copilot Chat, GitHub Copilot agent mode)
Edit .vscode/mcp.json for workspace, or run MCP: Open User Configuration for global:
{
"servers": {
"noticed": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@noticed/cli", "mcp"],
"env": { "NOTICED_API_KEY": "nk_live_…" }
}
}
}
Note VS Code uses servers (not mcpServers) and requires type.
Windsurf
Edit ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"noticed": {
"command": "npx",
"args": ["-y", "@noticed/cli", "mcp"],
"env": { "NOTICED_API_KEY": "nk_live_…" }
}
}
}
Cline (VS Code)
Edit ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json:
{
"mcpServers": {
"noticed": {
"command": "npx",
"args": ["-y", "@noticed/cli", "mcp"],
"env": { "NOTICED_API_KEY": "nk_live_…" }
}
}
}
Continue
MCP works in agent mode only. Add .continue/mcpServers/noticed.yaml:
mcpServers:
- name: noticed
command: npx
args: ["-y", "@noticed/cli", "mcp"]
env:
NOTICED_API_KEY: nk_live_…
Zed
Edit ~/.config/zed/settings.json (note: key is context_servers, not mcpServers):
{
"context_servers": {
"noticed": {
"command": "npx",
"args": ["-y", "@noticed/cli", "mcp"],
"env": { "NOTICED_API_KEY": "nk_live_…" }
}
}
}
Claude Code skill
This package ships a Claude Code skill at skills/noticed-search/SKILL.md. Skills don't auto-discover from node_modules — symlink it into your skills directory once:
# Personal scope (all projects)
mkdir -p ~/.claude/skills
ln -s "$(npm root -g)/@noticed/cli/skills/noticed-search" ~/.claude/skills/noticed-search
# Project scope (this repo only)
mkdir -p .claude/skills
ln -s "$(npm root)/@noticed/cli/skills/noticed-search" .claude/skills/noticed-search
Or install the whole thing as a Claude Code plugin (skill + MCP bundled):
git clone https://github.com/noticedso/cli ~/.claude/plugins/noticed
# restart Claude Code
Quick start (CLI)
# 1. Mint an API key at https://www.noticed.so/dashboard/api-keys
# 2. Configure credentials
noticed config --set-key nk_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
# 3. Search your network
noticed search "react developers"
# 4. Find the shortest path to a specific person
noticed path @sarahml
noticed path --li sarah-chen
# 5. JSON output for scripting
noticed search "frontend" --json | jq '.hits[].github_login'
Authentication
The CLI authenticates with a Bearer API key. Mint one at /dashboard/api-keys — the secret is shown once at create time, so copy it immediately. Tokens look like nk_live_…, are rate-limited to 60 requests per minute per key, and can be revoked from the same page.
CLI commands
noticed search <query>
Search your developer network for people, companies, skills, or topics.
noticed search "AI engineers" # natural language
noticed search "@sarahml" # GitHub username
noticed search "CTO at Vercel" # job title + company
noticed search "kubernetes" --source github # filter by source
noticed search "react" --limit 10 --json # paginated JSON output
noticed search "engineers" --csv > out.csv # CSV export
noticed search "Sarah Chen" --paths # also fetch paths to top hits
| Flag | Description | Default |
|---|---|---|
-l, --limit <n> |
Maximum results | 25 |
-o, --offset <n> |
Pagination offset | 0 |
-s, --source <src> |
Filter: github or linkedin |
all |
--sort <col:dir> |
Sort: name:asc, company:desc |
none |
-p, --paths |
Fetch shortest paths to the top 5 hits | off |
-j, --json |
Output raw JSON | off |
--csv |
Output as CSV | off |
--no-color |
Disable colors | auto |
noticed path [target]
Find the shortest connection path between you and a person.
noticed path @sarahml # by GitHub login
noticed path 12345 # by github_user_id
noticed path --li sarah-chen # by LinkedIn username
noticed path @sarahml --json # machine-readable
Logins are resolved via search before the path lookup. Pass --li to skip the search step and look up by LinkedIn username directly.
noticed config
noticed config # show current config
noticed config --set-url <url> # set API URL
noticed config --set-key <key> # set API key
noticed config --show # show current config
Config is stored at ~/.config/noticed/config.json (XDG-compliant).
noticed mcp
noticed mcp # start MCP server over stdio
noticed mcp --log-level debug # with debug logging on stderr
noticed completion <shell>
noticed completion bash >> ~/.bashrc
noticed completion zsh >> ~/.zshrc
noticed completion fish > ~/.config/fish/completions/noticed.fish
Environment variables
| Variable | Description | Required |
|---|---|---|
NOTICED_API_KEY |
API key minted at https://www.noticed.so/dashboard/api-keys | yes |
NOTICED_API_URL |
Override the noticed instance URL. Defaults to https://www.noticed.so — only set this if you self-host |
no |
NOTICED_BASE_URL |
Alias for NOTICED_API_URL |
no |
Precedence: CLI flags > environment variables > config file.
MCP tools
The MCP server exposes exactly two tools — both meta-tools that bridge to the noticed agent's capability registry. The client's LLM uses search to discover capabilities at runtime, then calls execute by name.
| Tool | Description |
|---|---|
search |
Discover noticed capabilities by keyword and optional category. Returns names, descriptions, categories, and JSON parameter schemas. Call with no arguments to list everything. |
execute |
Run a capability by exact name. Pass capability arguments in the args object. |
Example client-side flow:
// 1. Find the right capability
tools/call search { "query": "missions" }
// → returns [{ name: "list_missions", parameters: {…}, … }, …]
// 2. Run it
tools/call execute { "capability": "list_missions", "args": {} }
// → returns the user's missions
The ~50 chat-safe capabilities cover developer-network search (search_network, get_connection_path, my_profile, my_network, my_activity, …), missions/goals/milestones, PRM (people / interactions / stages), virtual filesystem and persona files, persistent memory, web search and fetch, and cron scheduling. Nine chat-only capabilities (in-chat messaging, referral invites, Cursor Cloud agents) are filtered server-side.
Test with the MCP Inspector:
# stdio
npx @modelcontextprotocol/inspector npx @noticed/cli mcp
# hosted (Streamable HTTP) — set Authorization: Bearer <key> in the inspector UI
npx @modelcontextprotocol/inspector
# URL: https://mcp.noticed.so/api/mcp
Or by hand against the stdio server:
echo '{"jsonrpc":"2.0","method":"initialize","id":1,"params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' | noticed mcp
echo '{"jsonrpc":"2.0","method":"tools/list","id":2}' | noticed mcp
Programmatic usage
import { NoticedApiClient } from "@noticed/cli";
const client = new NoticedApiClient({
baseUrl: "https://www.noticed.so",
apiKey: "nk_live_…",
});
const results = await client.search("AI engineers", { limit: 10 });
const path = await client.path({ to: results.hits[0]?.github_user_id });
console.log(results.hits, path);
Development
npm install
npm run build
npm test
npm run lint
npm run check-types
Self-hosting noticed
The CLI defaults to the hosted noticed instance at https://www.noticed.so. Self-hosting noticed itself is possible but operationally heavy — the value depends on a pre-ingested GitHub + LinkedIn collaboration graph (hundreds of GiB of ClickHouse data, daily GHArchive ingestion, paid LinkedIn API access, OpenAI + Anthropic keys, NextAuth OAuth apps). Most users want the hosted service.
If you do run your own instance, set NOTICED_API_URL to its URL — everything else works the same:
export NOTICED_API_URL=https://noticed.your-domain.com
export NOTICED_API_KEY=nk_live_…
noticed search "AI engineers"
The source for the hosted service lives at https://github.com/noticedso/noticed.
License
MIT © noticed
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.