AI Agent · MCP Server

Your AI agent, connected to AudFlo.

Connect Claude Code, Cursor, or any MCP-capable agent to AudFlo. It runs your audit, reads the blockers, applies the Fix Pack directly to your codebase, and verifies the fixes after you deploy.

Other tools tell your agent what is wrong. AudFlo hands it the fixes — and checks they shipped.

Connect my agent Get an API key

01.

Create an API key

02.

Add the server

Add https://audflo.com/api/mcp to your agent with the key as a Bearer header — copy-paste configs below.

03.

Ask your agent

Say “audit my site with AudFlo and fix what it finds.” The agent handles the rest, and asks before anything that costs a scan.

Connect in one command

Replace audflo_live_YOUR_KEY with your key from Settings.

Claude Code

claude mcp add --transport http audflo https://audflo.com/api/mcp \
  --header "Authorization: Bearer audflo_live_YOUR_KEY"

Cursor — .cursor/mcp.json

{
  "mcpServers": {
    "audflo": {
      "url": "https://audflo.com/api/mcp",
      "headers": {
        "Authorization": "Bearer audflo_live_YOUR_KEY"
      }
    }
  }
}

Claude Desktop & other stdio clients — via mcp-remote

{
  "mcpServers": {
    "audflo": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote", "https://audflo.com/api/mcp",
        "--header", "Authorization: Bearer audflo_live_YOUR_KEY"
      ]
    }
  }
}

Six tools, one loop: audit → fix → verify

run_audit

Audit any site on demand

Your agent starts a full AudFlo audit — 32 checks across four pillars — and gets a scan id back. It always asks you before spending a scan.

get_audit_results

Read the results, not a PDF

Score, rank, pillar breakdown, and top blockers with evidence — as structured data your agent can reason about directly.

get_fix_pack

Apply fixes to your codebase

Headlines, FAQ block with schema, meta descriptions, founder bio, robots.txt, and llms.txt — implementation-ready, so your agent edits your project files instead of handing you homework.

verify_published_fix

Confirm fixes actually shipped

After you deploy, your agent rechecks your live site against the audit and reports exactly which checks now pass.

list_audits

Pick up where you left off

Your audit history with scores and status, so your agent reuses an existing audit instead of running a new one.

check_ai_crawlers

Check AI crawler access — free

Robots policy plus live user-agent testing for GPTBot, ClaudeBot, PerplexityBot, and more. No API key needed.

Workflows as prompts

The server also ships ready-made workflows your agent can pick from its prompt menu — no instructions to write.

/fix-my-visibility

Audit the site, explain the blockers, apply the Fix Pack to your codebase, and set up verification — end to end.

/verify-my-fixes

Recheck your live site after a deploy and report what now passes, what still fails, and why.

/whats-changed

Diff your two most recent audits: score movement, pillar changes, blockers resolved or introduced.

FAQ

Agent & MCP questions

What is the AudFlo MCP server?+

It is an endpoint at https://audflo.com/api/mcp that speaks the Model Context Protocol, the open standard AI agents use to call external tools. Once connected, agents like Claude Code and Cursor can run AudFlo audits, read results, pull Fix Pack assets, and verify published fixes — inside the same session where they edit your code.

How do I get an API key?+

Sign in to AudFlo and open Settings → API Keys. Keys start with audflo_live_ and are shown once at creation — AudFlo stores only a hash. You can revoke a key at any time from the same screen.

Can my agent spend money or run audits without asking?+

No. Tools that consume a scan credit require an explicit confirmation argument, and the tool descriptions instruct agents to check with you first. Read-only tools like listing audits or fetching results run freely.

Which clients are supported?+

Any MCP client that supports remote servers over Streamable HTTP: Claude Code, Cursor, Windsurf, and others. Clients that only support local stdio servers, like Claude Desktop, connect through the standard mcp-remote bridge shown above.

How is this different from the web app?+

Same audits, same Fix Packs, same account. The difference is where the work happens: the web app shows you fixes to implement; the MCP server hands them to the agent that is already editing your codebase, which can apply them and then verify the deploy.

Put your agent to work.

Run a free audit first, or connect your agent and let it drive.

Run Free Audit Connect my agent