const agentDefinition = {
id: "find-food",
displayName: "Find Food",
publisher: "tanzeela",
version: "0.0.12",
model: "anthropic/claude-4-sonnet-20250522",
toolNames: [
"spawn_agents"
],
spawnableAgents: [
],
inputSchema: {
params: {
type: "object",
properties: {
location: {
type: "string",
description: "Location to search for restaurants (e.g., \"Mission District SF\", \"Downtown Portland\"). If omitted, location will be auto-detected."
},
dietaryRestrictions: {
type: "array",
items: {
type: "string"
},
default: [
"gluten-free",
"dairy-free",
"pescatarian"
],
description: "List of dietary restrictions (e.g., [\"gluten-free\", \"dairy-free\", \"pescatarian\"])"
}
}
},
prompt: {
type: "string",
description: "Additional context or preferences for restaurant search (optional)"
}
},
includeMessageHistory: true,
outputMode: "last_message",
spawnerPrompt: `Spawn when you need to find nearby restaurants`,
systemPrompt: ``,
instructionsPrompt: `
Use the Exa MCP to help find restaurants that meet the specified dietary restrictions within 1 mile of the specified location.
The dietary restrictions are provided in params.dietaryRestrictions as an array of strings (e.g., ["gluten-free", "dairy-free", "pescatarian"]).
If params.location is not provided, ask the user for their location before searching.
Return at most 5 restaurants in your final output. Pick the 5 best candidates that match the dietary restrictions and proximity criteria.
For each of the (up to 5) candidate restaurants, spawn a research-restaurant agent and pass the dietary restrictions to it.
Format each restaurant result CONCISELY using exactly this structure (and nothing else):
**<Restaurant Name>** — <distance from user, e.g. "0.4 mi" or "~8 min walk">
- Menu items: <comma-separated list of specific dishes that meet the dietary restrictions>
- Dietary accommodations: <short summary of how the restaurant accommodates the restrictions, e.g. "dedicated GF fryer; tamari available; 100% dairy-free">
Do NOT include "Perfect for", "Bonus", "Highlights", ratings, hours, addresses, ⭐ icons, or any other sections. Keep it tight.
`,
stepPrompt: ``,
mcpServers: {
exa: {
url: "https://mcp.exa.ai/mcp?",
type: "http",
params: {},
headers: {}
}
},
inheritParentSystemPrompt: false
}