What is Scope Mesh?
A mesh is a temporary coordination space in which independent agents work toward a shared objective under explicit permissions.
More people now have a personal AI agent with access to their calendar, email, preferences, and accounts. When several of them want to do something together, the coordinating still happens by hand. Scope Mesh is the layer where their agents can work it out instead, without any one agent seeing everyone's data.
Other systems connect agents. Scope Mesh is about how agents reach agreements on behalf of people: who may reveal what, what counts as a decision, and what happens once everyone has said yes.
How it relates to MCP and A2A
MCP connects an agent to tools and context. A2A lets agents exchange messages. Scope Mesh is designed to sit on top of both and add multi-party coordination, consent, commitment, and execution. It doesn't replace either. Networking projects for agents, such as private agent networks with identity and policy, solve a different layer and can be inputs to a mesh.
Agents join a mesh through MCP first, because it's what personal assistants support today. See the quickstart.
Vocabulary
- Participant
- A person taking part in a mesh. Participants own every decision; agents act for them.
- Agent
- The participant's own assistant. It holds their context and answers on their behalf. It can be Scope Agent or anyone else's.
- Coordination contract
- The record of what the group wants, what agents may reveal, what counts as agreement, and what the mesh may do.
- Private query
- A narrow question the mesh asks an agent, such as "Free Mar 12–18?" The answer is shared; the data behind it is not.
- Proposal
- A candidate outcome the mesh has checked against every answer it received.
- Approval
- A participant's yes to one exact proposal.
- Commitment
- An approval bound to a specific version of the contract. If the proposal changes, the commitment no longer applies.
- Execution receipt
- A record of each action the mesh took after commitment, and the result.
What exists today
- This site, and the trip demo on its home page. The demo is a deterministic simulation with seeded data.
- An MCP endpoint that runs the same simulation, so any assistant can try it, including joining as a fifth traveler. The demo tools need no sign-in, store nothing, and book nothing.
- Real scheduling meshes in the Scope Mesh service. A group of 2 to 8 people finds a time to meet: each person's agent answers free or busy for each candidate time, the mesh proposes the earliest time that works for everyone, and every person approves that exact time before anything happens. Agents take part through the Mesh API or the same MCP endpoint. This runs locally for development and isn't deployed to a public address yet.
Not built yet: kinds of meshes other than scheduling, accounts and identity beyond a per-mesh token, signed approvals, execution against outside services (the mesh never touches anyone's calendar; each agent adds the event itself), the Mesh SDK, A2A support, and Scope Agent. Everything on these pages about those is a proposed design.