Agent capabilities

Operating principle

Agents act inside the systems where risk lives. Write-back and execution are gated per product — see each agent entity page for approval boundaries.

Machine-readable surfaces

• OpenAPI specification — public X-ray and concierge endpoints
• /.well-known/agent.json — capability and trust card
• llms.txt — curated routing document
• sitemap.xml — canonical URL inventory

Safe public actions

• POST /api/scanScan a public company website for Business X-ray prefills
• POST /api/diagnoseScore diagnostic signals and recommend a Persidian agent
• POST /api/chatAsk short follow-up questions about a recommendation
• POST /api/reportsSave a consent-based shareable X-ray report (30-day retention)

These endpoints do not write to customer systems. Rate limits apply. See Trust and Security for boundaries.

Product entity pages

• Sikizana — canonical reference at https://persidian.com/agents/sikizana
• Nuncio — canonical reference at https://persidian.com/agents/nuncio
• Lenitnes — canonical reference at https://persidian.com/agents/lenitnes
• DataBard — canonical reference at https://persidian.com/agents/databard
• Weft — canonical reference at https://persidian.com/agents/weft
• Diversifi — canonical reference at https://persidian.com/agents/diversifi

Limitations

• No customer system write access through public endpoints
• Website scans inspect public pages only
• Product subdomains own detailed product UX; entity pages are canonical for discovery
• Pilot features are labeled on entity pages and must not be cited as mainnet production

MCP

A narrowly scoped MCP server for Persidian public operations is planned. Until then, use the OpenAPI spec above or call the REST endpoints directly. No MCP write tools are published for customer systems.