Agents with Cloudflare Clef

Local MCP server that gives AI agents structured decisions with probabilities from Clef-Flash through llama.cpp, with no network needed.

Uses Clef to decide whether a prompt needs a frontier model or a cheap one, behind an OpenAI-compatible proxy and library to cut spend.

A Pi extension that registers Clef decision models as tools so an agent can ask for odds instead of prose, routed through your own authenticated Cloudflare AI Gateway.

An experiment that lets Clef (Flash) operate a macOS desktop while keeping everything local.

A Pi extension that builds a neutral context and lets Jev or Clef return a probability distribution over choices, to advise the agent rather than command it.

An agent does the legwork while Clef makes each judgment call from discovery to regression.

Clef-Flash reviews each reasoning step of a llama-server model and triggers rollback or reflection.

CLI, MCP server and Chrome extension bridge that lets Clef operate your own signed-in browser session.

An MCP server that uses Workers AI Clef to judge page content so an agent can navigate a browser, adapted from the jev-browser engine.

A browser agent where Jev or Clef chooses each action and a vision LLM steps in only when unsure; ships as CLI, library and MCP server.

A shared local computer-use runtime that uses Clef with OmniParser, MCP and a CLI to turn a high-level goal into a bounded sequence of screen actions.

A Birthday Week demo that prechecks tool calls with Clef-flash and caches the decisions in Workers KV, showing cache hits and misses.

Cloudflare Worker prototype that asks Clef-flash whether a tool call is appropriate before the LLM makes it.

Interactive demo of Clef-Flash showing tool routing at 97% confidence, multi-question decisions and questions about video frames.

Replays a Pi agent log as a timeline and shows which step Clef scores as the fault, a debugging aid for agent runs.