The Model Context Protocol (MCP) is an open standard that enables large language model assistants to discover data tools and execute remote procedures securely. PaceCast operates a fully compliant MCP server supporting WebStandard Streamable HTTP and Server-Sent Events (SSE).
Protocol Architecture & Endpoints
| Endpoint / Resource | Type | HTTP Method, Transport & Purpose |
|---|---|---|
/api/mcp (or /mcp)required | POST / GET / SSE | Core MCP streamable endpoint. Accepts JSON-RPC 2.0 payloads for 'tools/list' and 'tools/call'. Authenticated via Bearer JWT. |
/.well-known/oauth-protected-resourcerequired | GET (Public) | RFC 9728 Protected Resource Metadata. Advertises supported scopes ('read:clients', 'read:campaigns') and documentation URIs. |
/.well-known/oauth-authorization-serverrequired | GET (Public) | RFC 8414 Authorization Server Metadata. Advertises PKCE S256 support and token endpoints. |
RFC Discovery & Metadata
Compliant agents discover PaceCast's authentication capabilities automatically by querying RFC 9728 metadata:
{
"resource": "https://pacecast.app",
"authorization_servers": ["https://pacecast.app"],
"scopes_supported": [
"read:clients",
"read:campaigns",
"write:campaigns",
"read:creatives",
"write:creatives",
"read:billing"
],
"bearer_methods_supported": ["header"],
"resource_documentation": "https://pacecast.app/docs"
}Configuring Claude & ChatGPT
To connect PaceCast to Claude Desktop or local MCP client agents, add the server to your MCP configuration file:
{
"mcpServers": {
"pacecast": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-fetch", "https://pacecast.app/api/mcp"],
"env": {
"PACECAST_API_TOKEN": "<YOUR_BEARER_JWT>"
}
}
}
}Core Tool Contracts
PaceCast exposes 28 distinct tools across accounts, diagnostics, creative intelligence, and strategy. Key tools include:
| Tool Name | Type | Parameters & Description |
|---|---|---|
list_clientsrequired | Read-Only | Returns authorized client workspaces accessible to the user with currency and timezone metadata. Arguments: {}. |
get_client_overviewrequired | Read-Only | Fetches connected platforms, latest analysis run freshness, and current blended CPA. Arguments: { client_id: string }. |
get_campaign_performancerequired | Read-Only | Queries daily spend, conversions, and CPA timeseries across Meta and Google Ads. Arguments: { client_id: string, start_date?: string, end_date?: string }. |
analyze_creative_fatiguerequired | Read-Only | Scans hooked assets for frequency decay and conversion rate degradation. Arguments: { client_id: string, threshold_days?: number }. |
get_creative_recommendationsrequired | Read-Only | Retrieves prioritized production briefs and 15-second creative hypotheses for video editors. Arguments: { client_id: string }. |
create_spend_scenariorequired | Analytical | Simulates budget scaling (Hold, Scale, Plan) with power-law diminishing return modeling. Arguments: { client_id: string, scenario_mode: string, budget_delta?: number }. |
Sample JSON-RPC Tool Execution
An AI assistant calls get_client_overview over the streamable transport:
{
"jsonrpc": "2.0",
"id": 104,
"method": "tools/call",
"params": {
"name": "get_client_overview",
"arguments": {
"client_id": "01934c2b-e104-7da5-b982-b7e61a6c4290"
}
}
}The PaceCast server responds with structured performance artifacts and metadata:
{
"jsonrpc": "2.0",
"id": 104,
"result": {
"client": {
"id": "01934c2b-e104-7da5-b982-b7e61a6c4290",
"name": "NorthStar Athletics",
"currency": "USD"
},
"connected_platforms": [
{ "platform": "meta", "primary_kpi": "cpa" },
{ "platform": "google", "primary_kpi": "roas" }
],
"latest_analysis": {
"status": "completed",
"blended_cpa": 14.82,
"forecast_wape": 0.114,
"naive_baseline_wape": 0.187
}
}
}