Vauda
Sign inStart free
NATIVE MCP SERVER

Your library, as context for any AI.

Vauda exposes your catalog through Model Context Protocol — the open standard spoken by Claude, ChatGPT and Cursor. The MCP server is a thin adapter over the same API the app uses, so an agent sees your assets with exactly the same permissions as a user. No exporting, no duplicating, no moving files.

Vauda MCP
C
Claude
G
ChatGPT
>_
Cursor
tool: search_assets("ejecutivos…")
tool: get_asset(id: 0xa72…)
✓ 14 results · 280ms
How to connect

Ready in three steps.

1 · Authorize
From your workspace, generate an MCP server URL and authorize the client with OAuth 2.1 (DCR + PKCE). The token is bound to an identity, with its scopes and organization permissions.
2 · Connect
Paste the URL into Claude, ChatGPT or Cursor. The client discovers the available tools automatically — nothing to install, no code to write.
3 · Work
Ask your AI in natural language: search, cite, tag or upload. Every action respects RBAC and usage rights, and lands in the audit log.
Reference

Available tools.

The server exposes a small, auditable set of tools. All of them inherit the token's RBAC, respect usage rights and return signed URLs instead of binaries.
search_assets
Hybrid search (semantic + keywords + filters) over the catalog the token is allowed to see. Returns scored results with previews.
get_asset
An asset's full metadata (AI catalog, EXIF/IPTC, rights) plus a signed preview or download URL — never the full binary.
list_collections
Lists the collections, folders and smart collections visible to the token, to scope searches or uploads.
tag_asset
Adds or edits an asset's tags and keywords. Subject to the token's role (requires editor or above).
upload_asset
Uploads a new file into a collection and triggers the cataloging pipeline. Requires upload permission.
get_usage_rights
Checks an asset's usage rights (country and expiry) before using it, so the agent respects restrictions.
Security & control

Power for your AI, control for you.

Exposing your library to an agent doesn't mean losing governance. Every security layer in Vauda applies the same over MCP as over the web.

OAuth 2.1 (DCR + PKCE)
Standard, secure authorization. No shared keys in plain text; every client registers dynamically and uses PKCE.
RBAC inheritance
The token lives inside an organization and a role. An agent never sees more than that user would see in the app.
Signed URLs + previews
Results never include the full binary: only metadata and signed, expiring links.
Audit log per token
Every call — which tool, which asset, what result — is recorded and reviewable by administrators.
Rate limiting per token
Per-token limits protect the platform and keep each connected agent's consumption in check.
Rights enforcement
Country and expiry restrictions apply over MCP too: an expired or out-of-territory asset is not delivered.
Use cases
Briefs & research
“Claude, show me last quarter's aerial forest shots.” The agent searches, cites the assets and builds the moodboard — with links that open in Vauda.
Conversational ingest
“Upload this set and tag it by product.” The agent creates the collection, uploads the files and lets the pipeline catalog them.
Audit & cleanup
“Find assets with expired rights, or duplicates.” The agent lists them respecting permissions and proposes actions — you approve.

Connect your library to AI.

The MCP server is included from the Pro plan up. Start free and connect it to Claude, ChatGPT or Cursor in minutes.