Overview
The Whether Agent Interface provides structured, machine-readable access to regime assessments for AI agents, automation tools, and Model Context Protocol (MCP) integrations.
Endpoint
GET /api/agent?cadence={cadence}
https://whether.fyi
Parameters
Time horizon for the summary. Options: weekly, monthly, quarterly, yearly
Response Schema
{
cadence: "weekly" | "monthly" | "quarterly" | "yearly";
defaultCadence: "weekly";
supportedCadences: string[];
summary: {
copy: string;
provenance: string[];
recordDateLabel: string;
};
agentHandoff: {
payload: AgentPayload;
prompt: string;
};
summaryHash: string | null;
links: {
self: string;
llms: string;
discovery: string;
};
generatedAt: string;
version: "v1";
}
Agent Payload Structure
The agentHandoff.payload contains structured regime data:
{
report_date: string;
fetched_at: string;
source: string;
regime: {
key: string;
label: string;
description: string;
};
scores: {
tightness: number;
riskAppetite: number;
};
thresholds: {
tightness: { threshold: number; exceeded: boolean };
riskAppetite: { threshold: number; exceeded: boolean };
};
constraints: string[];
data_warnings: string[];
signals: Array<{
id: string;
label: string;
value: number | null;
unit: string;
record_date: string;
fetched_at: string;
source_url: string;
}>;
macro_signals: Array<{
id: string;
label: string;
value: number | null;
unit: string;
record_date: string;
fetched_at: string;
source_url: string;
}>;
skills: Array<{
id: string;
label: string;
purpose: string;
inputs: string[];
outputs: string[];
}>;
pm_handoff: {
decision_focus: string;
briefing_actions: string[];
};
}
Example Requests
cURL
JavaScript
Python
TypeScript SDK
curl https://whether.fyi/api/agent?cadence=weekly
const response = await fetch(
"https://whether.fyi/api/agent?cadence=weekly",
{
headers: { accept: "application/json" },
cache: "no-store"
}
);
const data = await response.json();
console.log(data.agentHandoff.payload.regime);
import requests
response = requests.get(
"https://whether.fyi/api/agent",
params={"cadence": "weekly"},
headers={"accept": "application/json"}
)
data = response.json()
regime = data["agentHandoff"]["payload"]["regime"]
print(f"Current regime: {regime['label']}")
import { pullRecentSiteInfo } from "whether/lib/agentInterface";
const info = await pullRecentSiteInfo({
cadence: "weekly",
siteBaseUrl: "https://whether.fyi"
});
console.log(info.payload.agentHandoff.payload.regime);
Model Context Protocol (MCP)
Whether provides an MCP server for seamless integration with Claude Desktop and other MCP-compatible tools.
get_agent_brief
Fetch a cadence-specific Whether summary with structured agent handoff payload and prompt.
Parameters:
cadence (optional): "weekly" | "monthly" | "quarterly" | "yearly"
Returns: Complete agent interface response with regime data.
list_agent_skills
List Whether agent skills expected by the agent handoff payload.
Parameters: None
Returns: Array of skill definitions with inputs and outputs.
pull_recent_site_info
Pull the latest Whether site agent brief from /api/agent for fresh site context.
Parameters:
cadence (optional): Time horizon for the summarysiteUrl (optional): Custom base URL (defaults to production)
Returns: Site URL, endpoint, cadence, and full payload.
MCP Server Setup
Install dependencies
bun install @modelcontextprotocol/sdk
Configure Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:{
"mcpServers": {
"whether": {
"command": "bun",
"args": ["run", "/path/to/whether/scripts/mcp/whether-mcp-server.ts"]
}
}
}
Restart Claude Desktop
The Whether MCP tools will be available in the tool picker.
MCP Server Implementation
The Whether MCP server is implemented in scripts/mcp/whether-mcp-server.ts:
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { buildAgentInterfaceResponse } from "../../lib/agentInterface";
const server = new McpServer({
name: "whether-agent-interface",
version: "1.0.0",
});
server.tool(
"get_agent_brief",
"Fetch a cadence-specific Whether summary",
{ cadence: z.enum(["weekly", "monthly", "quarterly", "yearly"]).optional() },
async ({ cadence }) => {
const response = await buildAgentInterfaceResponse(cadence ?? "weekly");
return { content: [{ type: "text", text: JSON.stringify(response, null, 2) }] };
}
);
const transport = new StdioServerTransport();
await server.connect(transport);
Agent Skills
The agent payload includes skill definitions that guide downstream automation:
1. Regime Summary
Purpose: Explain the market climate in plain English with top macro signals.
Inputs: regime, scores, macro_signals, data_warnings
Outputs: summary_bullets
2. Constraint Translation
Purpose: Turn constraints into clear guardrails for roadmap, hiring, and pricing.
Inputs: constraints, thresholds, signals
Outputs: constraint_brief
3. PM Question Log
Purpose: Surface open questions the PM must answer before acting.
Inputs: constraints, signals, macro_signals
Outputs: pm_questions
Agent Prompt
The agentHandoff.prompt field provides guidance for autonomous PM assistants:
You are an autonomous PM assistant.
Follow the skills list in the JSON payload and deliver outputs in the same order.
Use the JSON payload to draft:
1) A 3-bullet summary of the market climate.
2) The top 3 constraints that should shape roadmap decisions.
3) A short list of questions to confirm with the PM before acting.
Context: Whether Report for 2025-02-28 (CONSTRAINED_CAUTIOUS).
Discovery
Whether exposes machine-readable metadata at standard discovery endpoints:
- LLMs.txt:
/llms.txt - Site metadata for AI agents
- Agent Discovery:
/.well-known/whether-agent.json - API capabilities
Rate Limiting
The agent API is designed for periodic polling:
- Recommended interval: 1 hour for
weekly cadence
- Data freshes weekly with new Treasury reports
- Use
summaryHash to detect content changes
CORS
The agent API includes CORS headers for browser-based integrations:
Access-Control-Allow-Origin: *
Access-Control-Allow-Methods: GET, OPTIONS