The agent harness for better falafel.
falafel.bot is the orchestration layer that turns frontier models into capable falafel agents. Its compute-efficient harness resolves your craving, coordinates restaurant search and menu tools, and optimizes every pita across taste, distance, value, and dietary preferences.
falafel.bot connects frontier models to discovery, personalization, and restaurant infrastructure through purpose-built scaffolding for the full falafel journey.
Describe what you want in natural language. The harness coordinates model reasoning, specialized tools, and preference memory to return a personalized recommendation in seconds.
A research interface for testing tool-use flows, comparing shop rankings, and inspecting the orchestration behind every recommendation.
A standard interface connecting falafel agents to restaurant catalogs, menu graphs, availability signals, and transaction-ready ordering tools.
Private research preview
Evaluate the agent harness against real falafel tasks, inspect tool-use traces, and compare model routing strategies. Playground access is opening to a limited research cohort.
$ inspect falafel.harness [ready] frontier model routing [ready] restaurant discovery tools [ready] preference graph orchestration [ready] compute-efficient execution Private preview access is required.
Model Chickpea Protocol
MCP gives agents a standard way to discover shops, inspect menus, evaluate live availability, and prepare an order. The protocol turns fragmented restaurant data into a reliable tool layer for autonomous falafel decisions.
Sister system
thefalafel.ai is the full falafel intelligence platform — frontier models, craving detection, and restaurant infrastructure in one system. Explore the research, the architecture, and the roadmap.
Explore thefalafel.ai ↗Private preview
We are onboarding an initial cohort of users, restaurant partners, and developers. Join the waitlist for preview access, product updates, and upcoming protocol documentation.