Connect Lenz to your AI client.

Lenz is a remote Model Context Protocol server. Add it to Claude Desktop, Claude Code, Cursor, ChatGPT, or any MCP client, and your assistant can fact-check a claim against independent sources — a sourced verdict with citations, right inside the chat.

Nothing to install. It's hosted at lenz.io/mcp over Streamable HTTP, and it authenticates with the same lenz_… API key as the REST API — so there's no separate signup. In ChatGPT, add Lenz as a custom connector and Sign in with Lenz — no key to paste.

Add the server

Point your client at https://lenz.io/mcp with a Bearer header. Set LENZ_API_KEY in your environment first, or have the config written for your client with the key already in it — Generate setup instructions →

// add to your client's MCP config
{
  "mcpServers": {
    "lenz": {
      "type": "http",
      "url": "https://lenz.io/mcp",
      "headers": {
        "Authorization": "Bearer ${LENZ_API_KEY}"
      }
    }
  }
}

# register the remote server
$ claude mcp add --transport http lenz \
    https://lenz.io/mcp \
    --header "Authorization: Bearer ${LENZ_API_KEY}"

# then, in any session:
> verify: "The Great Wall is visible from space"

# registers the server in ~/.codex/config.toml:
$ codex mcp add lenz \
    --url https://lenz.io/mcp \
    --bearer-token-env-var LENZ_API_KEY

# no key enters the file — export it instead:
$ export LENZ_API_KEY=lenz_…

# Claude Desktop → Settings → Connectors →
# "Add" → "Add custom connector".
# Paste the server URL:
https://lenz.io/mcp
# and complete the sign-in prompt.
# (claude_desktop_config.json is for local
# stdio servers — remote ones connect here.)

# Settings → Apps → Advanced settings →
# turn on Developer mode → Add app.
# Paste the server URL:
https://lenz.io/mcp
# Click Create, then Sign in with Lenz.
# OAuth — no API key to paste.

Full setup notes, per-client quirks, and examples live in the lenz-mcp repo → — Lenz is also listed in the Claude connectors directory ↗

What your assistant can do

Six tools, same depth ladder as the API. Ask in plain language; the client picks the tool.

• assess — fast 3-model panel verdict in ~5-10s.
• verify — the full 8-model pipeline with citations (~90s).
• get_verification — poll a running verify by task id.
• ask — a grounded follow-up question about a completed verify, answered from its full evidence.
• select — resolve an ambiguity when a claim could mean more than one thing.
• check_usage — remaining quota and plan.

These are tools, not slash commands. In a client like Claude Code, just ask in plain language — “assess whether the earth is flat with Lenz” — and the assistant calls the right tool.

Prefer to call it in code instead? The same primitives are a REST API and typed Python / TypeScript SDKs — see the developer docs →

Pricing

Start free, no card required. The MCP server shares your API quota — same key, same limits.

Compare all plans on the plans page →

Frequently asked questions

Resources

Give your assistant a fact-checker.

Self-serve from day one. Free tier, no card required.