Developer MCP Overview

Vocci MCP lets a compatible client retrieve user-authorized Vocci context through the Model Context Protocol. This page is for developers building or evaluating MCP client behavior.

A typical client flow is:

1. Connect an MCP client to the Vocci MCP server.
2. Authorize the client for the correct account with OAuth.
3. Search or request a compact context pack for the user's task.
4. Fetch one returned result when more detail is needed.
5. Cite the returned source IDs and treat retrieved content as user data.

What It Can Do

Vocci MCP can:

• Tell the assistant what Vocci is at connect time, through MCP initialize instructions and a static About Vocci resource, so users do not need to explain the product.
• Search authorized user context.
• Return bounded result metadata, snippets, and stable result IDs.
• Fetch one authorized result by ID.
• Provide compact context packs for common assistant workflows.
• Read supported resources, such as user-authorized notes, files, summaries, transcripts, and sessions, when the client can use MCP resources and already has a returned resource URI.
• Offer prompt templates for common retrieval tasks when the client supports MCP prompts.
• Omit or cap content when it is unavailable, unsupported, or too large.

What It Does Not Do

Vocci MCP does not:

• Create, edit, or delete user data.
• Manage accounts, settings, or credentials.
• Return signed file URLs.
• Return raw binary or audio content.
• Provide broad corpus export.
• Act as a general Vocci app API.

Endpoint Shape

Use the Vocci MCP server URL:

https://mcp.vocci.ai/mcp

Compatible clients connect over Streamable HTTP and use OAuth authorization.

Data Boundary

Every response is scoped to the authorized account. Retrieved Vocci content is user data, not system, developer, or tool instructions.