Ask an AI assistant

Sherlock has an MCP server. Connect Claude Code, the Claude apps, Cursor, VS Code, Codex, or any other MCP client to it, and ask about your logs, traces, and metrics in plain language. Settings → MCP in Sherlock shows the same instructions with copy buttons.

Try asking

Tell me about the state of the system using the Sherlock MCP tools
How many errors did we get in the last hour?
Group today’s errors and tell me which one is new
Which of today’s errors actually matter, and which are just noise?
Latency spiked at 14:20. Check the metrics, then pull the traces behind it

Connect with OAuth

This is the recommended way. Point your client at the server URL and sign in through the browser. There is no token to copy, store, or rotate. Access follows your Sherlock account and can be revoked at any time. Server URL: https://mcp.sherlocklabs.dev/mcp

Set up your client

Add the server, with no header and no token:

claude mcp add sherlock -t http "https://mcp.sherlocklabs.dev/mcp"

Run /mcp in Claude Code, select sherlock, then choose Authenticate.
Your browser opens. Sign in with your Sherlock account and approve access. Claude Code stores the credentials for you.

Press ⌘/Ctrl Shift J to open Cursor Settings.
Select Tools & Integrations, then New MCP Server.

Add Sherlock to the config that opens:

{
  "mcpServers": {
    "sherlock": {
      "url": "https://mcp.sherlocklabs.dev/mcp"
    }
  }
}

Click Needs login next to Sherlock and complete sign-in in the browser.
Optional. To use Sherlock from cursor-agent, run cursor-agent mcp login sherlock.

From your terminal, add the server:

codex mcp add sherlock --url https://mcp.sherlocklabs.dev/mcp

The next time you run codex, Sherlock is available and opens the OAuth flow to connect to your account.Or edit ~/.codex/config.toml and add the remote MCP configuration, then restart any running codex session:

[mcp_servers.sherlock]
url = "https://mcp.sherlocklabs.dev/mcp"

Any client that speaks MCP over Streamable HTTP with OAuth 2.1 can connect with the server URL. Sherlock advertises its authorization server at /.well-known/oauth-protected-resource, and that server supports dynamic client registration. Most clients discover everything they need from the URL alone, with no client id or secret to configure.

Connect with an API token

A long-lived token works where the browser sign-in flow does not: headless machines and CI, shared automation, or clients that do not implement OAuth. The token grants the same access as your account. Treat it like a password.

claude mcp add sherlock -t http "https://mcp.sherlocklabs.dev/mcp" \
  --header "x-api-key: <your-token>"

Generate a token with Generate MCP token on Settings → MCP. Sherlock inserts it into the command on that page. Copy it right away. You cannot see it again. You can also use an existing token from the API Tokens page.

Ask about these docs

The documentation has its own MCP server, separate from the one above. Connect it and the agent answers from these pages while you integrate. See Docs for AI agents.

Get started

Send your first data before you ask about it.

Traces

What the assistant sees when it pulls a trace.

Send data

Node.js SDK

Go SDK

Ask an AI assistant

Try asking

Connect with OAuth

Set up your client

Connect with an API token

Ask about these docs

Get started

Traces

​Try asking

​Connect with OAuth

​Set up your client

​Connect with an API token

​Ask about these docs

​Related topics

Get started

Traces

Try asking

Connect with OAuth

Set up your client

Connect with an API token

Ask about these docs

Related topics