Dossier MCP server
{
"mcpServers": {
"imboard-dossier": {
"command": "npx",
"args": [
"-y",
"@ai-dossier/mcp-server@latest"
]
}
}
}{
"mcpServers": {
"imboard-dossier": {
"command": "npx",
"args": [
"-y",
"@ai-dossier/mcp-server@latest"
]
}
}
}{
"mcpServers": {
"imboard-dossier": {
"command": "npx",
"args": [
"-y",
"@ai-dossier/mcp-server@latest"
]
}
}
}{
"servers": {
"imboard-dossier": {
"command": "npx",
"args": [
"-y",
"@ai-dossier/mcp-server@latest"
]
}
}
}{
"mcpServers": {
"imboard-dossier": {
"command": "npx",
"args": [
"-y",
"@ai-dossier/mcp-server@latest"
]
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y @ai-dossier/mcp-serverversion 1.0.4
What it does
This MCP server enables LLMs to discover, verify, and execute dossiers — portable, signed skill sets written in Markdown with JSON frontmatter. It is for developers who want to share reusable AI workflows across different LLM tools with built-in integrity verification.
- Discover and run shared dossiers (Markdown-based skill workflows) from a registry or URL without manual copy-pasting into each LLM tool.
- Verify the cryptographic signature and checksum of a dossier before an LLM executes it, ensuring authenticity and preventing tampering.
- Execute a versioned dossier workflow intelligently across different LLM platforms (Claude, ChatGPT, Cursor) with consistent behavior and documented success criteria.
Tools this server exposes
| Tool | Does |
|---|---|
| discover | List available dossiers from the registry or local store. |
| verify | Check the cryptographic signature and checksum of a dossier to confirm authenticity and integrity. |
| execute | Run a dossier workflow, interpreting its Markdown instructions and validating success criteria. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.