About Vantage
Vantage is a personal signal store for people doing AI-native work, helping them save and query observations gathered across the web. Its MCP server makes those saved signals available across compatible AI clients, with hosted and MIT-licensed self-hosted options.
Key features
- 01Save posts, articles, and discussions
- 02Full-text search with filters
- 03Manage sources, topics, and collections
- 04Analyze signals against business context
- 05Save analyses and view summaries
Use cases
- 01Keep research observations from social posts or articles
- 02Find saved signals by topic or source
- 03Compare business strategy with collected signals
- 04Reuse research across AI tools
MCP support
Vantage provides 13 MCP tools for compatible clients such as Claude, Claude Code, and Cursor to capture, search, and analyze signals. It supports stdio and HTTP.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://vantage.cavendo.ai/mcpclaude mcp add --transport http vantage https://vantage.cavendo.ai/mcp{"mcpServers":{"vantage":{"type":"http","url":"https://vantage.cavendo.ai/mcp"}}}Open in Cursor{"servers":{"vantage":{"type":"http","url":"https://vantage.cavendo.ai/mcp"}}}Open in VS Code{"mcpServers":{"vantage":{"type":"http","url":"https://vantage.cavendo.ai/mcp"}}}{"mcpServers":{"vantage":{"serverUrl":"https://vantage.cavendo.ai/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://vantage.cavendo.ai/mcpAuth requiredTools
Details
From the project READMEVantage
Track what you're seeing. Use it later.
Vantage is a personal signal store for AI-native work — a place to capture and query the things you notice across the web.
A post on X. A thread on LinkedIn. An insight from an article. A competitor move. Instead of losing it across tabs, screenshots, and notes, you save it — and use it later.
One memory layer across all your AI tools
Your AI tools don't share memory.
- Claude doesn't know what you saw on X
- ChatGPT doesn't know what you read yesterday
- Your code tools don't know your research
Vantage fixes that.
Save signals once, and access them anywhere you have an MCP-compatible client — across Claude, Cursor, and other supported tools.
How it works
- Connect Vantage to your AI (via MCP)
- Save signals directly from your workflow
- Ask questions on top of what you've saved
Example:
"Save this as a signal"
"What have I saved about pricing?"
"Show me signals from Lemkin"
What Vantage is (and isn't)
- Not a scraper — you choose what matters
- Not a feed — no noise or algorithmic clutter
- Not a knowledge base — no system to maintain
Vantage is a simple way to keep track of what you're seeing — so your AI can actually use it later.
Hosted + open source
- Free hosted version: vantage.cavendo.ai
- Self-hosted: MIT licensed, run it anywhere
Quick Start
Option 1: Use the hosted version (fastest)
Sign up at vantage.cavendo.ai. Connect to Claude or another supported client. Start capturing.
Takes about 2 minutes. No install required.
Option 2: Run locally
npm install
npm run dev
# If 3020 is already in use, choose another port
PORT=YOUR_PORT npm run dev
# Generate your first API key
curl -X POST http://localhost:3020/api/auth/keys \
-H 'Content-Type: application/json' \
-d '{"name": "my-key"}'
The first API key can be created from localhost without auth. If you're bootstrapping a fresh remote deployment, set VANTAGE_BOOTSTRAP_TOKEN on the server and send the same value in the X-Vantage-Bootstrap-Token header on that first request.
The API key is returned as data.key in the JSON response.
Capture your first signal:
curl -X POST http://localhost:3020/api/signals \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer vtg_YOUR_KEY_HERE' \
-d '{
"signalType": "post",
"platform": "x",
"content": "Every SaaS company will need an agent layer within 18 months.",
"topics": ["AI Agents", "SaaS"],
"importance": "high"
}'
MCP Server
Vantage includes an MCP server with 13 tools for use with Claude, Claude Code, Cursor, or any MCP-compatible client.
If you want to use the bundled stdio MCP server directly, install its dependencies too:
cd mcp-server
npm install
cd ..
Claude Code
claude mcp add vantage --transport http "http://localhost:3020/mcp" --header "Authorization: Bearer vtg_YOUR_KEY"
For manual testing, note that /mcp is a streamable HTTP endpoint. MCP clients handle the response format automatically, but ad hoc curl calls need to be prepared for stream-oriented responses rather than a simple JSON-only request/response flow.
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"vantage": {
"command": "node",
"args": ["/path/to/vantage/mcp-server/src/index.js"],
"env": {
"VANTAGE_API_URL": "http://localhost:3020",
"VANTAGE_API_KEY": "vtg_YOUR_KEY_HERE"
}
}
}
}
MCP Tools
| Tool | What it does |
|---|---|
capture_signal | Save a post, article, quote, or discussion. Auto-creates sources and topics. |
search_signals | Full-text search with filters — platform, topic, date range, importance. |
list_recent_signals | Latest signals feed. |
get_signal | Full detail on a specific signal. |
manage_source | Create or update a person/account you're tracking. |
list_sources | See everyone you're tracking with signal counts. |
manage_topic | Create or update a topic/theme. |
list_topics | All topics with signal counts. |
manage_collection | Group signals into named collections. |
set_business_context | Upload strategy docs for comparison. |
analyze_signals | Assemble signals + context for analysis. |
get_dashboard | Summary — counts, top topics, top sources, highlights. |
save_analysis | Persist an analysis for later reference. |
Integrations
- OpenClaw community skill: vantage-shared-intel — capture signals, search prior intel, and generate shared digests against the same Vantage dataset from OpenClaw.
REST API
Most API requests require Authorization: Bearer vtg_YOUR_KEY.
Exceptions:
GET /healthPOST /api/auth/keysfor first-key bootstrap from localhost, or from a remote client that supplies a validX-Vantage-Bootstrap-Token
| Method | Endpoint | Description |
|---|---|---|
| POST | /api/signals | Create a signal |
| GET | /api/signals | List signals with filters |
| GET | /api/signals/search?q=query | Full-text search |
| GET | /api/signals/feed | Chronological feed |
| GET | /api/signals/:id | Signal detail |
| POST | /api/sources | Create a source |
| GET | /api/sources | List sources |
| POST | /api/topics | Create a topic |
| GET | /api/topics | List topics |
| POST | /api/collections | Create a collection |
| POST | /api/contexts | Upload business context |
| GET | /api/dashboard/summary | Dashboard summary |
| POST | /api/auth/keys | Generate API key |
Tech Stack
| Component | Technology |
|---|---|
| Runtime | Node.js 20+ / Express / ESM |
| Database | SQLite via better-sqlite3 |
| Validation | Zod |
| Auth | API key (vtg_ prefix), SHA-256 hashed |
| MCP | @modelcontextprotocol/sdk (stdio + HTTP) |
Environment Variables
| Variable | Default | Description |
|---|---|---|
PORT | 3020 | Server port |
NODE_ENV | development | Environment |
DATABASE_PATH | ./data/vantage.db | SQLite file path |
CORS_ORIGIN | * | CORS allowed origin |
VANTAGE_BOOTSTRAP_TOKEN | unset | Allows first API key creation from a non-localhost client when sent as X-Vantage-Bootstrap-Token |
License
MIT — use it however you want.
Built by Jonathan Arehart at Cavendo AI.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.