We gate our own blog. Scored 17 October 2026 with the same engine you get, glossary rules, balanced strictness.
Dokeo ships an MCP server (Model Context Protocol) so Claude Desktop, Cursor, and any MCP client can score + fix content without leaving the chat.
Exposed tools (15)
scan_contentScore inline content against 19 content checks. Returns a verdict, scores, and action items.
scan_urlFetch a public URL and score it. Same shape as scan_content.
batch_scanScore several drafts in one call and return per-item verdicts.
suggest_fixesScore content and return the fix text for checks that did not pass.
list_content_typesList the 17 content types and the extra checks each one adds.
statsAudit-log totals: evaluations, verdict breakdown, and average scores.
search_knowledge_baseSearch brand-voice guides, playbooks, and benchmarks in the knowledge base.
check_brand_voiceCheck text against the brand-voice banned-phrase list.
ask_dokeoAsk a question grounded in this workspace's scan data. scope=growth is operator-only.
list_custom_rulesList this tenant's custom quality rules.
submit_objectSubmit a non-content object. Content drafts should use scan_content.
run_testRun the test adapter for an object. Errors when that type has no adapter connected.
get_issuesList open issues from tests. Filter by severity, status, or object.
submit_fixRecord a fix for an issue. It does not re-test; call run_test again to verify.
request_approvalMove an object into the approval queue for a person to sign off.
HTTP transport (cloud clients)
For Claude.ai web, Cursor cloud, and other clients that don't speak stdio, Dokeo exposes JSON-RPC 2.0 over HTTP:
curl -X POST https://api.dokeo.co/api/v1/mcp/jsonrpc -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "scan_content",
"arguments": {
"content": "your draft goes here",
"content_type": "blog"
}
}
}'Discover available tools with GET /api/v1/mcp/info (no auth).
Setup with Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on your OS and add:
{
"mcpServers": {
"dokeo": {
"command": "uvx",
"args": ["--from", "dokeo", "dokeo-mcp-server"],
"env": {
"DOKEO_API_KEY": "dk_your_tenant_key",
"DOKEO_API_URL": "https://api.dokeo.co"
}
}
}
}Restart Claude Desktop. The dokeo tools appear in the chat under "Tools".
Setup with Cursor
Edit ~/.cursor/mcp.json:
{
"mcpServers": {
"dokeo": {
"command": "uvx",
"args": ["--from", "dokeo", "dokeo-mcp-server"],
"env": {
"DOKEO_API_KEY": "dk_your_tenant_key",
"DOKEO_API_URL": "https://api.dokeo.co"
}
}
}
}Use it
# From Claude Desktop or Cursor, just ask:
"Score my latest draft for SEO + AEO + GEO"
# Claude will call the MCP tool automatically and show you:
# - VERDICT (BLOCKER / FAIL / WARNING / OPTIMIZATION / PASS)
# - score per check
# - the offending sentences (inline_issues)
# - concrete fix suggestions
#
# Claude can then auto-rewrite the draft and re-score until it passes.