AI assistants
AI assistants and coding tools that support the Model Context Protocol (MCP) can connect to CalmLens and work in one of your projects: moderate text or a URL, look up assets and decisions, explain a result, and organise folders. Add CalmLens as a remote MCP server with the URL below, then approve the connection in your browser.
https://api.calmlens.com/mcpConnect an assistant
Add the URL wherever your assistant lists connectors or MCP servers. Assistants with a configuration file take an entry like this:
{
"mcpServers": {
"calmlens": {
"url": "https://api.calmlens.com/mcp"
}
}
}The first time the assistant uses CalmLens, your browser opens CalmLens to approve it. Any app can register under any name, so the page shows where approving sends you: only continue if you started connecting the app there. Choose the project and what the assistant can do:
| Access | The assistant can |
|---|---|
| Read only | List, search and open assets, explain decisions, and read folders and events. |
| Read and change | Also moderate new content, and change, move or delete assets and folders. |
An assistant never manages members, API keys, connectors, billing or project settings, and never shares assets publicly or creates download links for them, whatever your own role allows.
How access works
Approving a connection creates an API key for the assistant in the project, named after the assistant. It acts as you: events record its changes as made by the assistant for you, and it can only do what you can. If your role in the project changes so you no longer hold what you approved, or you leave the project, the assistant stops working until you connect it again.
Assistant keys don't expire. Your Portal › Profile lists the assistants you connected in every project, and you can disconnect any of them there whatever your role. Anyone who manages a project's keys can also revoke them under Portal › API keys. Its access can't be edited; revoke it and connect again to choose differently.
Assets hold content from your users. Treat what an assistant reads from them as untrusted: CalmLens tells assistants not to follow instructions found in asset content, and read-only access limits what a misled assistant could change.
Tools
Each tool runs one API endpoint with the assistant's key, so the same validation, limits and usage apply. projectId defaults to the connected project.
| Tool | Endpoint |
|---|---|
create_asset | POST / |
list_assets, search_assets, get_asset, list_asset_children | GET /, GET /, GET /, GET / |
explain_asset | POST / |
update_asset, move_asset, retry_asset | PATCH /, POST /, POST / |
delete_asset, delete_asset_content | DELETE /, DELETE / |
list_folders, get_folder, create_folder, update_folder, move_folder, delete_folder | GET /, GET /, POST /, PATCH /, POST /, DELETE / |
list_events | GET / |
Assistants mark the delete tools as destructive and usually ask before running them. A read-only connection isn't offered the tools that change things.
Command line
The calmlens CLI signs in with calmlens login: it shows a code to enter in CalmLens in your browser, and keeps the key for later commands. See Local development.