MCP Server splicewire · v1.3.0

A curated Model Context Protocol server that gives AI clients first-class access to your Splicewire knowledge base and content builder — hybrid keyword + vector search over your silos, structured outline generation, and export. Purpose-built tools, not an auto-wrapped REST surface.

Endpoint

MCP https://{tenant}.app.splicewire.com/mcp

Replace {tenant} with your subdomain. The server is tenant-scoped: it operates on the knowledge base for that tenant only.

Authentication

The endpoint speaks the MCP streamable-HTTP transport and authenticates over OAuth 2.0 — clients discover the authorization server automatically and prompt you to sign in. No manual token wiring is required for OAuth-capable clients.

Connect a client

For Claude Desktop or Cursor, add the server to your MCP config (using mcp-remote to bridge OAuth):

{
  "mcpServers": {
    "splicewire": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://{tenant}.app.splicewire.com/mcp"
      ]
    }
  }
}

Clients with native remote-MCP + OAuth support can point straight at the endpoint URL without the mcp-remote bridge.

Tools (20)

Search

search

Search the Splicewire RAG knowledge base. Uses hybrid Reciprocal Rank Fusion (RRF) combining PostgreSQL full-text keyword search with pgvector semantic similarity. Falls back to keyword-only if vector search is unavailable. This is the primary search tool.

List Silos

list_silos

List the knowledge base silos (categories/collections). Silos organize fragments into hierarchical groups. Use this to discover what content areas are available.

Get Fragment

get_fragment

Get the full details of a specific fragment by its ID. Use this after searching to retrieve the complete text content of a fragment.

List Fragments

list_fragments

List fragments in the knowledge base, optionally filtered by silos or tags. Use this to browse available content.

Describe Image

describe_image

Describe an image the user uploaded, understand the art style, and use it as an inspiration for constructing an appropriate prompt that will be later used to generate a skybox.

Generate Image

generate_image

Create an image as described and return it as a markdown image link.

You Tube Video Info

youtube_video_info

Get information about a YouTube video, including captions.

List Outlines

list_outlines

List content outlines. Returns a paginated list of outlines with title, description, item count, and ID.

Get Outline

get_outline

Get a content outline with its full hierarchical item tree. Returns outline metadata and all items with their content summaries.

Create Outline

create_outline

Create a new content outline. After creating, use generate_outline to generate the hierarchical structure from the description.

Update Outline

update_outline

Update a content outline's title or description.

Delete Outline

delete_outline

Delete a content outline and all its items. Requires confirm: true to execute. Call without confirm first to see what will be deleted.

Generate Outline

generate_outline

Generate the hierarchical structure for a content outline using AI. The outline must have a description and an associated chat. Creates sections/items based on the description. This runs synchronously and may take 30-60 seconds.

Add Outline Item

add_outline_item

Add a new item to a content outline. Items are hierarchical — use parent_id to nest items under an existing item.

Update Outline Item

update_outline_item

Update an existing content outline item. Can change summary, data (title, subtitle, body, image_prompt), parent, or order.

Delete Outline Item

delete_outline_item

Delete a content outline item and its descendants. Requires confirm: true to execute. Call without confirm first to see what will be deleted.

Generate Outline Item Content

generate_outline_item_content

Generate full content for a specific outline item using AI. Produces title, subtitle, body text, and image prompt. Runs synchronously and may take 30-60 seconds.

Generate All Outline Content

generate_all_outline_content

Queue content generation for ALL items in an outline. Items are processed sequentially in the background (30-60 seconds each). Returns immediately — use `get_outline` to check progress (items show [no content] or [N chars]).

Export Outline

export_outline

Export a content outline as markdown, HTML, or raw JSON. Use after generating content to get the finished document.

Resolve Json Schema

resolve_json_schema

Resolve and register a JSON Schema from a URI. Fetches the schema, stores it in the tenant database, and makes it available as a circuit schema ref or instructions schema.