MCP Integration
Your AI agent can sign its own outputs. In one config change.
MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude call external tools. The Verbitas MCP server exposes vb_sign, vb_verify, and vb_lookup as tools Claude can call in any conversation.
Install in 3 steps
1 Install the MCP server
npm install -g verbitas-mcp
2 Initialize (prompts for API key)
verbitas-mcp init
3 Add to Claude Desktop config (~/.claude/config.json)
{
"mcpServers": {
"verbitas": {
"command": "verbitas-mcp",
"env": { "VERBITAS_API_KEY": "your-key-here" }
}
}
} Available tools
| Tool | Description | Required params |
|---|---|---|
| vb_sign | Sign a file with a recipe | file_path, recipe |
| vb_verify | Verify provenance of a file | file_path |
| vb_lookup | Look up by perceptual fingerprint | file_path |
| vb_recipe_create | Create a custom recipe | yaml (admin scope) |
| vb_recipes_list | List available preset recipes | none |
| vb_audit_log | Get recent audit events | from, to (admin scope) |
Example conversation
User
Generate an image of a mountain and sign it with the genai recipe.
Claude
[calls vb_sign with file_path="mountain.jpg", recipe="image-genai-v1"] Done. The image has been signed with a C2PA manifest and TrustMark watermark. Manifest: https://verify.verbitas.io/m/01HXQ...
Safety notes
- — MCP tools only sign files at explicitly caller-provided paths — no arbitrary file system access.
- — The API key is read from environment variables only — never passed as a tool argument.
- — All API key scopes are configurable per-tenant.
AI agents that sign their own outputs.
Install verbitas-mcp. Configure Claude. Done.