Model Context Protocol

Bring AIDE into your editor

Cursor, Claude Desktop, and any other MCP client can launch AIDE scans, read scores, and surface remediation tasks right beside your code. Same API key, same 195 checks, zero context-switching.

Create an API key Upgrade to Pro

MCP is enabled on Pro and up

Free accounts cannot connect to MCP. Once you upgrade, the same API key authenticates both REST and MCP traffic.

See pricing

Three steps to connect

1. Get an API key

From Account > API keys create a key starting with `aide_live_…`. The secret is shown once — store it somewhere safe.

Create an API key

2. Wire it into your MCP client

Add the server config below to Cursor, Claude Desktop, or your own agent. Send the API key in the Authorization header.

Cursor configuration · ~/.cursor/mcp.json

{
  "mcpServers": {
    "aide": {
      "url": "https://mcp.aide.tr/mcp",
      "headers": {
        "Authorization": "Bearer aide_live_yourkey_here"
      }
    }
  }
}

Claude Desktop configuration · macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "aide": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.aide.tr/mcp"],
      "env": {
        "AUTH_HEADER": "Bearer aide_live_yourkey_here"
      }
    }
  }
}

Verify the connection with cURL

curl https://mcp.aide.tr/mcp \
  -H "Authorization: Bearer aide_live_yourkey_here" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

3. Ask and ship

Once connected the agent gets `scan_url`, `get_scan`, `get_leaderboard`, and friends as tools. Try a natural prompt like "scan aide.tr in AIDE and summarise the three worst issues".

Tools you can call

The 27 tools we expose over MCP — each bound to the same permissions as the REST API.

scan_url

Starts a fresh AIDE scan for a URL; profile and SEO hints are optional.

get_scan

Returns score, badge, per-check results and recommendations for an existing scan id.

wait_for_scan

Blocks until a scan finishes, then returns the final summary.

list_scans

Lists your account’s recent scans, filterable by status and domain.

get_leaderboard

Returns the top N domains from the AIDE leaderboard, optionally by sector.

compare_scans

Diffs two scans across score, category and failing checks.

get_check_details

Gives one check’s weight, category and Learn link.

explain_issue

Explains one failing check with a Turkish and English fix.

get_project_context

Reads the stored project memory for a verified domain: business overview, audience, competitors, key pages, research log.

update_project_context

Writes project memory for a verified domain — partial update, with merge fields for competitors, key pages and the research log.

save_report

Saves a report you wrote as a self-contained HTML document to the account, and returns a link to open it.

list_reports

Lists the reports saved in the account, newest first.

whoami

Reports the key’s owner, plan, scopes, verified domains and remaining monthly scan quota.

list_domains

Lists the account’s domains with verification state and latest score.

list_checks

Lists the check registry, narrowable by profile or category.

list_profiles

Returns every scan profile with its check count and categories.

get_profile_info

Explains one profile in depth: category weights, severity mix, heaviest checks.

get_scan_issues

Returns a finished scan’s failing checks as one flat, severity-ordered list.

get_scan_diff

Compares a scan with the domain’s previous one: score, category and check deltas.

list_schedules

Lists a domain’s scheduled scans with cadence and next run time.

create_schedule

Sets up a recurring scan for a verified domain — each run spends one scan from the quota.

delete_schedule

Deletes a scheduled scan; its history stays, no further runs happen.

list_alert_rules

Lists the score-drop and regression alert rules.

get_alert_history

Returns recently fired alerts with what changed.

get_backlinks

Summarises inbound links AIDE observed in its own crawl (coverage is limited to it).

search_learn

Searches the Learn library by keyword, check id or tag.

get_learn_article

Fetches one Learn article in full, by slug.

Agent Skills

Plugins — install AIDE as a skill set

Instead of calling tools one by one, install AIDE as ready-made workflows. The plugin ships five skills, each driving the MCP server with its own discipline: audit, fix, compare, monitor and report.

Claude Code

/plugin marketplace add aide-tr/skills
/plugin install aide@aide-tr

Add the marketplace first, then install the plugin. Skills load at the start of a session.

Cursor

npx skills add aide-tr/skills

Cursor reads `.cursor-plugin/plugin.json` from the repository. To wire the MCP server by hand instead, the `~/.cursor/mcp.json` sample above is enough.

Codex

npx skills add aide-tr/skills

Codex reads `.codex-plugin/plugin.json` from the repository; the MCP definition is identical in both manifests.

Other agents (generic pattern)

npx skills add aide-tr/skills

Any client that supports the Agent Skills format uses the same command. Clients that do not can copy the server definition from `plugins/aide/mcp.json` directly.

The public GitHub mirror will be announced here once it is live; the planned address is: aide-tr/skills

Skills in the package

aide-audit
Scans a domain, ranks findings by severity × per-profile weight, verifies them against the live site, and returns a short action list with one "do this week" item.
aide-fix-issue
Takes a check id or issue title and writes the concrete fix for your stack (robots.txt, llms.txt, JSON-LD, meta, headers), then re-scans to confirm it passes.
aide-compare
Scores you against two to four competitors on the same profile, prices the gaps in points, and places you on the sector leaderboard.
aide-monitor
Reads a domain's scan history, diffs adjacent runs and names the check behind every score move. Scheduled scans and alerts stay in the account panel.
aide-report
Writes a self-contained one-page HTML report from scan results — Turkish or English, no external resources.

Skills use the same API key: export `AIDE_API_KEY` with your `aide_live_…` key before starting your agent. MCP access is available on Pro and above.

Frequently asked questions

How is security handled?

Every request is signed with `Authorization: Bearer aide_…`. The key is bound to your account; quotas + scopes are enforced server-side. If a key leaks, rotate it from Account > API keys.

What are the limits?

Free does not include API keys. Pro: unlimited manual scans + 5,000 API calls per month (3 keys). Business: unlimited manual scans + 50,000 API calls (15 keys). Enterprise: custom. 429s should be retried with exponential backoff.

Is OAuth supported?

OAuth 2.1 client registration (per MCP 2025-06-18) is on the roadmap. For now MCP only accepts Bearer API keys.