Skip to main content
Monnet is an agentic decision facilitator that runs in the cloud. While it excels at decision making, it can help with anything you do with your team: bug fixes, brainstorming ideas, planning a team event, researching a market, and more. This guide covers the core lifecycle, built-in capabilities, and tips for configuring Monnet. For step-by-step walk-throughs, see the Product section.

​
The motion lifecyle

Monnet’s atomic unit is a motion: a natural language proposal for action. Every non-trivial thing your team decides (a feature to ship, a contract to sign, a plan to commit to) can be captured as a motion. When you give Monnet a motion, it works through three phases with your team: gather input, align, and take action. These phases blend together. Monnet uses tools throughout, messaging individuals, posting comments, prompting agents, making updates to the motion, or requesting approval. Motion lifecyle diagram The lifecycle adapts to what you ask. A motion with a clear proposal might only need approvals. A small bug fix will move to execution directly. A complex decision might go through several iterations of deliberation. Monnet decides what each step requires based on what it learned on the previous step, chaining dozens of actions together and course-correcting along the way. You and your team are a part of this loop. You can see what Monnet is planning to do and you can steer it in a different direction. Monnet works autonomously but stays responsive to your input. The motion lifecycle is powered by three components: the team, tools and user interfaces. Monnet serves as a multiplayer harness around an LLM: it provides the tools, context management, and interfaces that turn a language model into a capable facilitator agent.

​
The team

A team in Monnet is comprised of humans and agents. Humans have roles at the workspace level (Owner / Admin / Member) and can also have motion-level roles (Editor / Commenter / External). See the Glossary for the permission matrix. Agents are AI teammates. They can pick up a motion, do the work, and report back. You add one through a connector. The connector available today is the Local Runner (Claude Code) — Monnet runs Claude Code on your own machine, using your subscription, repos, and permissions. Managed agents (hosted by Anthropic) and GitHub Actions connectors are on the way. Agents are similar to humans: they appear in your team list, get assigned work, and report back in the motion.

​
Tools

Tools are what make Monnet capable of coordination. Without tools, an LLM can only answer with text. With tools, Monnet can read, write, message, and delegate. Monnet’s built-in tools fall into four categories: Monnet picks which tools to use based on the motion and what it learned along the way. When you say “let’s discuss the impact of Opus 4.7 on us”, Monnet might search the web, draft a motion, ask you follow-up questions, loop in the most impacted teammates, identify next steps, request approval, and kick off new motions for those next steps.

​
Interfaces

Interfaces are the way you and your team interact with Monnet. The three primary surfaces:
  • Inbox — the motions that need your input, across all your workspaces. Your home page.
  • Motion page — the proposal itself: body, plan, discussion, versions, files. Where you comment, approve, and chat with Monnet.
  • Workspace — team, agents, policy (MONNET.md), skills, settings. The admin surface for your organization.

​
MONNET.md

Every workspace has a MONNET.md file. It’s the brief Monnet reads at the start of every motion — how your team works, what your priorities are, who approves what. You write it in plain English. MONNET.md grows with your team’s shared understanding. Add a line when Monnet makes the same mistake twice, when you find yourself re-typing the same clarification, or when a new teammate would need the same context to be productive.
See the Product guide for how to edit it.