redCacti MCP server
A read-only Model Context Protocol server that lets AI writing tools pull internal-link suggestions from your redCacti crawl. Point any MCP client at the endpoint below and authenticate with an API key.
https://api.redcacti.com/mcp X-API-Key: rc_live_….
Create a key in the
Connect your AI
wizard.
Tools
Four read-only tools. Nothing is written back.
analyze_article
read:analysis Analyze a draft and return ranked internal-link placements.
| Parameter | Type | Required | Description |
|---|---|---|---|
title | string | required | The article title (1–512 chars). |
content | string | required | The full article body. |
site_id | string | optional | Target site UUID (from list_sites). |
content_format | text | html | markdown | default markdown | Format of the body. |
max_suggestions | integer 1–30 | default 10 | Max placements to return. |
min_similarity | number 0.5–1.0 | default 0.65 | Minimum semantic similarity. |
min_confidence | number 0.3–1.0 | default 0.50 | Minimum placement confidence. |
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "analyze_article",
"arguments": {
"title": "How to audit internal links",
"content": "Internal links are the connective tissue of your site...",
"content_format": "markdown",
"max_suggestions": 5
}
}
} list_sites
read:sites List the account’s sites with ids, urls, and domains. Call this first.
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": { "name": "list_sites", "arguments": {} }
} get_site
read:sites Get one site’s detail: pages_crawled, health_score, is_crawling.
| Parameter | Type | Required | Description |
|---|---|---|---|
site_id | string | required | The site UUID (from list_sites). |
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": { "name": "get_site", "arguments": { "site_id": "<uuid>" } }
} get_crawl_status
read:crawls Poll a crawl job’s status and page counts.
| Parameter | Type | Required | Description |
|---|---|---|---|
crawl_id | string | required | The crawl job UUID. |
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": { "name": "get_crawl_status", "arguments": { "crawl_id": "<uuid>" } }
} Worked example: draft in, placements out
You are writing a post and want internal links to pages you already have. Send the
draft to analyze_article; it returns
ranked placements, each with the anchor text to use and the character offsets where
it belongs.
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "analyze_article",
"arguments": {
"title": "Internal linking for SaaS",
"content": "Your content architecture decides how well search engines discover pages...",
"content_format": "markdown",
"max_suggestions": 3
}
}
} {
"jsonrpc": "2.0",
"id": 1,
"result": {
"content": [{
"type": "text",
"text": "{"suggestions":[{"target_page_url":"...",
"suggested_anchor_text":"content architecture",
"anchor_start_pos":42,"anchor_end_pos":61,
"impact_level":"High", ...}]}"
}],
"isError": false
}
} Limits
- Read-only. No tool writes to your site, pages, or account.
- Rate limit: 60 requests/minute per principal (key id, else user id). This is protection only, with no call quotas.
- Free plan bounds: 1 site and up to 1,000 pages per site.
- Scopes: a key only reaches tools its scopes allow. Missing scope →
403.
Troubleshooting
The key is wrong, incomplete, or revoked. Copy the full rc_live_… value and confirm it is still active on the API keys page.
The key lacks the scope a tool needs. Create a new key with read:sites, read:crawls, and read:analysis.
Too many requests in a minute. Wait for the Retry-After window and slow down. Do not retry in a tight loop.
analyze_article needs a crawled site. Check pages crawled is greater than zero in the dashboard and run a crawl first.
The draft may not overlap your existing pages, or thresholds are too strict. Lower min_similarity / min_confidence, or write more on-topic content.
Set it up in a minute
The wizard creates a scoped key, generates your client config, and tests the connection.
Connect your AI