Give your AI a document brain.
VATIS is a Model Context Protocol server for EU document governance. Any AI — Claude, ChatGPT, Gemini, Cursor — can propose artefacts, mint traceable UDIS identifiers and apply policy, without ever touching your tenant credentials.
{
"name": "vatis.issue_udis",
"arguments": {
"track": "MWP",
"artefact": "SITE",
"ado": "F3391"
}
}One MCP server. Every model that matters.
Native MCP client. Add VATIS in Claude Desktop settings.
Connect VATIS as a custom connector in ChatGPT.
Reach VATIS tools over the MCP HTTP transport.
Drop the VATIS server into ~/.cursor/mcp.json.
Expose UDIS lookups inside VS Code agent mode.
Streamable HTTP · JSON-RPC 2.0 · OAuth 2.1.
The protocol your AI already speaks.
Model Context Protocol is the open standard for giving AI models tools and data. VATIS implements it end-to-end — so any compliant client can discover the same governed operations over your SharePoint, Graph and Azure DevOps estate.
- › vatis.propose_artefact
- › vatis.issue_udis
- › vatis.resolve_udis
- › vatis.apply_policy
- › vatis.audit_query
// GET /v1 → tools/list
{
"tools": [
{
"name": "vatis.propose_artefact",
"description":
"Match a plain-language brief to an approved artefact template.",
"inputSchema": {
"type": "object",
"properties": {
"text": { "type": "string" },
"actor": { "type": "string", "format": "email" }
},
"required": ["text","actor"]
}
},
{
"name": "vatis.issue_udis",
"description":
"Reserve a Unified Document Identifier bound to an ADO work item."
},
{
"name": "vatis.resolve_udis",
"description":
"Return provenance, policy and location for an existing UDIS."
}
]
}Identifiers your auditors can read out loud.
Every UDIS is four segments: what programme it serves, what kind of artefact it is, its position in that sequence, and the ADO work item that authorised its birth.
Five layers, one round trip.
From the moment your AI describes an intent to the moment the site, page or template is provisioned and identified.
- 01 · INTENT
AI describes what the user wants
A model connected via MCP calls vatis.propose_artefact with a plain-language brief. VATIS answers with a matched template, permissions, and a reserved identifier — before anything is provisioned.
// MCP tool call from the AI { "method": "tools/call", "params": { "name": "vatis.propose_artefact", "arguments": { "text": "Project site for Hornsea 4 EIA reviews", "actor": "leonard.h@documente.eu" } } } - 02 · MATCH
Approved patterns, not free-form
VATIS pins the request to a governed template — retention, classification and naming baked in. No shadow sites.
GET /vatis/templates?fit=team-site → { id: "MWP-TEAM-V3", permissions: "owner/member/visitor", metadata: ["project","track","phase"] } - 03 · PROVISION
Graph & SharePoint, executed cleanly
The MCP server issues the actual Microsoft Graph calls with a service identity — the AI never touches your tenant credentials.
POST https://graph.microsoft.com/v1.0/sites { "displayName": "Hornsea 4 — EIA", "template": "MWP-TEAM-V3" } - 04 · UDIS
Every artefact gets a traceable ID
A Unified Document Identifier is minted and bound to the ADO work item that authorised it. Provenance is queryable forever.
POST /udis/issue { "track":"MWP","artefact":"SITE","ado":"F3391" } → "MWP-SITE-00042-F3391" - 05 · POLICY
Enforced at provisioning time
Naming, classification and retention are applied by VATIS — not a checklist your AI must remember.
applyPolicy({ naming: "mwp-baseline", classification:"internal", retention: "7y-project" })
Point your AI at mcp.documente.eu.
Copy the snippet for your client. VATIS handles OAuth 2.1, dynamic client registration and audit logging.
// ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"vatis": {
"url": "https://mcp.documente.eu/v1",
"transport": "http"
}
}
}