Quickstart
Connect Claude, Claude Code, Cursor or any MCP client to the Skylit MCP server, with Skylit sign-in or an API key.
The endpoint is https://mcp.skylit.ai/mcp. There are two ways to authenticate:
| How | Best for | |
|---|---|---|
| Sign in with Skylit (OAuth) | Add only the URL. Your client opens a Skylit consent page; click Approve. | Claude, Claude Code, Cursor, VS Code, ChatGPT |
| API key | Send Authorization: Bearer <your-api-key> | Headless agents, scripts, clients without OAuth |
Either way the connection uses your account's credits and limits. OAuth connections show up on the Developer page as "Connected app: ..." and can be disconnected there.
Claude (desktop and claude.ai)
Settings, Connectors, Add custom connector. Name it Skylit, set the URL to
https://mcp.skylit.ai/mcp, click Connect, then approve on the Skylit page.
No key or config file is needed.
Using an API key instead (config file)
Bridge to the remote endpoint with mcp-remote
in claude_desktop_config.json, then restart Claude Desktop:
{
"mcpServers": {
"skylit": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.skylit.ai/mcp",
"--header",
"Authorization: Bearer YOUR_API_KEY"
]
}
}
}Claude Code
claude mcp add --transport http skylit https://mcp.skylit.ai/mcpThen run /mcp, pick skylit and choose Authenticate. For a headless
machine, use a key instead:
claude mcp add --transport http skylit https://mcp.skylit.ai/mcp \
--header "Authorization: Bearer $SKYLIT_API_KEY"Cursor
Add the server to ~/.cursor/mcp.json (global) or .cursor/mcp.json in a
project, then click Connect (or Login) next to skylit in Cursor
Settings, MCP:
{
"mcpServers": {
"skylit": { "url": "https://mcp.skylit.ai/mcp" }
}
}To use a key instead, add a header:
{
"mcpServers": {
"skylit": {
"url": "https://mcp.skylit.ai/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}Codex (OpenAI)
Add the server to ~/.codex/config.toml and give Codex your key through an environment variable:
[mcp_servers.skylit]
url = "https://mcp.skylit.ai/mcp"
bearer_token_env_var = "SKYLIT_API_KEY"export SKYLIT_API_KEY="your_api_key" # from app.skylit.ai/developer
codexCodex sends it as Authorization: Bearer <key>. The MCP server accepts the key only in that header, not as X-API-Key.
To sign in with your Skylit account instead of a key, add the server without bearer_token_env_var and run codex mcp login skylit. Some Codex versions need their remote MCP client feature turned on first; the key method above works on all of them.
"Unauthorized" when Codex connects means no credential reached the server. Check that SKYLIT_API_KEY is exported in the shell you start Codex from, then restart Codex.
ChatGPT
ChatGPT connects custom MCP servers in developer mode (available on paid plans):
Settings, Apps & Connectors, Advanced settings, turn on Developer mode,
then Create a connector with the URL https://mcp.skylit.ai/mcp and
OAuth authentication. ChatGPT opens the Skylit consent page; click Approve.
Other MCP clients
Any client that supports remote (streamable HTTP) MCP servers can use the URL
with OAuth, or the URL plus the Authorization header.
Try it
Once connected, ask a question that maps to a tool:
Were there any unusual bullish sweeps on TSLA today?
The agent will typically call flow_search to confirm the ticker, then sweeps
(or unusual_volume) and summarize. Browse example prompts
for more.
Raw HTTP (for developers)
The server speaks standard MCP over streamable HTTP and is stateless: there is no
Mcp-Session-Id, so every request stands alone and carries the key. Responses
may arrive as text/event-stream, so send an Accept header that includes it.
BASE=https://mcp.skylit.ai/mcp
AUTH="Authorization: Bearer $SKYLIT_API_KEY"
H=(-H "$AUTH" -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream")
# list tools
curl -sS "$BASE" "${H[@]}" -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
# call a tool
curl -sS "$BASE" "${H[@]}" -d '{"jsonrpc":"2.0","id":2,"method":"tools/call",
"params":{"name":"heat_levels","arguments":{"symbols":"SPY"}}}'Last updated