# mcpfiles.ai

> The filing cabinet for all your AI work. Store, organize, and search artifacts from AI conversations in one place.

mcpfiles.ai is an artifact storage platform designed for AI agents and their users. Every important conversation, decision, code snippet, or research finding can be summarized and filed here so nothing gets lost across ChatGPT, Claude, Gemini, Cursor, or any other AI tool.

## For AI Agents

mcpfiles.ai provides a Model Context Protocol (MCP) server that allows you to directly create, read, update, and search artifacts on behalf of your user.

### How to Connect

Your user will provide you with:
1. Their mcpfiles.ai subdomain (e.g., `chad.mcpfiles.ai`)
2. A time-limited access token starting with `mcpf_`

### MCP Server Configuration

```json
{
  "mcpServers": {
    "mcpfiles": {
      "url": "https://<subdomain>.mcpfiles.ai/mcp",
      "headers": {
        "Authorization": "Bearer <access_token>"
      }
    }
  }
}
```

### Available MCP Tools

When connected, you have access to these tools (depending on token permissions):

**Write permission (default):**
- `create_artifact` - Create a markdown, JSON, link, or skill artifact in a project
- `update_artifact` - Update an existing artifact (creates a new version)
- `add_tags` / `remove_tags` - Manage tags on artifacts

**Read permission:**
- `list_projects` - List all accessible projects
- `get_project` - Get project details
- `list_artifacts` - List artifacts with filtering
- `search_artifacts` - Full-text search across all artifacts
- `read_artifact` - Read artifact content
- `list_versions` / `read_version` / `diff_versions` - Version history

**Delete permission (requires CRUD token):**
- `delete_artifact` - Permanently delete an artifact

### Artifact Types

- `markdown` - Documents, notes, conversation summaries, meeting notes
- `json` - Structured data, API responses, configuration snapshots
- `link` - URLs to shared AI conversations (the system attempts to fetch and store the content)
- `skill` - Reusable prompt templates and instructions (markdown with YAML frontmatter)

### Typical Workflow

At the end of a conversation, your user may ask you to summarize the conversation and upload it. Here's how:

1. Ask the user for their mcpfiles.ai access token (or they may provide it upfront)
2. Use `list_projects` to find the right project
3. Use `create_artifact` with type "markdown", a descriptive title, relevant tags, and your summary as the content

### Token Permissions

Tokens are time-limited and scoped:
- **Duration**: 1 minute, 1 hour, 1 day, 7 days, or 30 days
- **Permissions**: read, write (default), readwrite, or full crud
- **Scope**: specific projects or all projects

Write-only tokens (the default) can only create new artifacts. They cannot read existing artifacts or delete anything. This is intentional for security.

## For Users

- **Sign up**: https://mcpfiles.ai/register/
- **Documentation**: See the dashboard after signing in
- **Terms**: https://mcpfiles.ai/terms/
- **Privacy**: https://mcpfiles.ai/privacy/
- **Contact**: hello@mcpfiles.ai

## Technical Details

- REST API at `https://<subdomain>.mcpfiles.ai/api/v1/`
- MCP server at `https://<subdomain>.mcpfiles.ai/mcp` (Streamable HTTP transport)
- Authentication via Bearer token in the Authorization header
- All artifact content is stored per-tenant in isolated databases
- Full version history on every artifact