Meeting coordination for the agentic web
Ask My Envoy is building meeting coordination for the agentic web, using MCP and A2A to let generalist AI agents work with specialized meeting-coordination agents.
AI agents are becoming capable of handling increasingly broad tasks: researching information, drafting content, working with business software, calling APIs, and taking actions on behalf of users. But the agentic web will not be built by one general-purpose AI agent reproducing every workflow in every domain.
Some problems require domain-specific logic, persistent rules, specialized integrations, and their own trust model. In those cases, a generalist AI agent needs to collaborate with a specialized agent, sometimes called a vertical AI agent: an agent designed to solve a narrower class of problems deeply rather than solve every problem broadly.
Meeting coordination is one of those domains. Ask My Envoy is building Envoys, specialized AI agents for meeting coordination. An Envoy is designed to represent a person's scheduling preferences, calendars, availability, trust rules, and coordination constraints, and to work with other people or agents to reach a meeting outcome.
Our premise is simple: when a generalist AI agent needs to arrange a meeting, it should not have to recreate meeting coordination from scratch. It should be able to work with an Envoy that already understands how that person's time can be coordinated.
Why general-purpose agents need specialized agents
A general-purpose AI agent can understand intent very well. It can understand a request such as “schedule a meeting with Paul next week,” identify that a calendar is involved, and potentially call an API that creates an event.
That does not mean it understands meeting coordination.
Creating a calendar event is a relatively simple software operation. A system needs a date, a start time, a duration, some participants, and access to a calendar. The difficult part usually happens before the event can be created.
The system has to determine which participants matter, which calendars are authoritative, what availability can be disclosed, what scheduling preferences apply, which time zones are involved, whether several people need to be consulted, and what happens when the first proposed time does not work.
Across organizations, that problem becomes more complex because the agent is no longer operating inside a single trusted environment. It may be interacting with people it does not know, calendars it cannot directly access, and other agents acting on behalf of different users.
That creates a trust problem as much as a scheduling problem.
An AI agent that can access someone's calendar is touching one of that person's most sensitive operational resources: their time. It may be able to infer availability, create commitments, move meetings, invite other people, or expose information about when someone is free.
But in cross-organization meeting coordination, trust is not only about what one agent can see in one calendar. It is also about the relationship between the people involved.
A meeting often brings together individuals from different companies, with different calendars, different policies, and different levels of trust. Some may already have an Envoy. Others may not. Some may be willing to expose availability directly, while others may only want to share the minimum information required to coordinate a meeting. In every case, the coordination agent needs enough information to do its job without assuming that every participant, calendar, or agent should be treated equally.
That makes trust relational.
The question is not simply whether an AI agent can access a calendar. It is whether the people involved trust one another enough to let their agents coordinate, what each side is allowed to disclose, and which actions each agent is authorized to take on behalf of the person it represents.
On the agentic web, that human relationship will increasingly be translated into relationships between AI agents. A person may trust a colleague, a client, or a partner to request time, while applying stricter rules to an unknown external agent. An Envoy therefore has to understand not only calendar permissions, but also who is asking, on whose behalf, and under what level of trust.
This is why Ask My Envoy treats trust as part of meeting coordination rather than as a separate security feature. A scheduling agent that simply reads and writes calendar events is not enough. A meeting coordination agent needs to understand participants, relationships, permissions, and boundaries across organizations.
That is the role of an Envoy.
An Envoy is a vertical AI agent specialized in meeting coordination. It represents a person's scheduling preferences, calendars, availability, and trust rules, and uses that context to coordinate with other people and agents while revealing no more information than necessary.
What is an Envoy?
An Envoy is a specialized AI agent for meeting coordination.
It sits between a person's calendars, scheduling preferences, trust rules, and the outside world. Its role is not simply to expose free time or create calendar events. Its role is to coordinate a meeting on that person's behalf while respecting the boundaries that person has defined.
An Envoy can work with a human, a general-purpose AI assistant, or another specialized agent. It can participate in a multi-turn coordination process where the requirements change, several people need to agree, and different levels of trust apply to different participants.
That makes an Envoy different from a booking link, a calendar API, or a generic scheduling tool. Those tools expose functions. An Envoy represents a coordination capability.
This is the model we are building toward for the agentic web: generalist agents handle the broad user relationship, while specialized or vertical agents handle domain-specific work. For meeting coordination, the specialized agent is an Envoy.
Open protocols are how agents find and work with each other
If generalist AI agents are going to work with specialized or vertical AI agents, they need standard ways to discover capabilities and communicate across systems. That is why Ask My Envoy is adopting open agent interoperability protocols rather than building a closed integration model.
MCP, the Model Context Protocol, gives AI hosts and agents a standard way to discover and connect to external capabilities. Ask My Envoy operates a production remote MCP server and is published in the official MCP Registry as:
com.askmyenvoy/meeting-scheduling-coordination
The production MCP endpoint is:
https://mcp.askmyenvoy.com/mcp
For human-readable context, this /developer page is the canonical source. The Registry entry and MCP endpoint are machine-facing infrastructure.
Ask My Envoy also supports A2A, the Agent2Agent protocol, for communication between agents. MCP and A2A solve different parts of the interoperability problem: MCP exposes capabilities to AI hosts and agents, while A2A provides a protocol for agents to communicate with other agents.
That distinction matters for meeting coordination because the interaction is not always a single tool call. A request can evolve as participants, timing, constraints, and trust relationships change. In those cases, a generalist AI agent needs more than access to a calendar function; it needs a way to work with a specialized meeting-coordination agent.
Ask My Envoy is building that agent layer around the Envoy: a vertical AI agent specialized in meeting coordination, designed to work with humans, generalist agents, and other agents across organizational boundaries.
The coordination logic stays with the Envoy
The protocols are only the external interface. The difficult part remains the coordination engine behind them.
An Envoy still has to deal with participant resolution, Google Calendar and Microsoft Calendar access, scheduling logic, time zones, trust and permissions, cross-organization boundaries, and the operational details required to turn a request into a confirmed meeting.
That is why we do not see agent interoperability as a replacement for the scheduling system. It is the way other agents gain access to it.
The objective is to let an external AI agent delegate meeting coordination without needing to understand how Ask My Envoy handles calendars, trust, scheduling constraints, or organizational boundaries internally.
Why this matters
Much of the current agent ecosystem is still organized around tools. An agent discovers a tool, sends it structured input, and receives a result. That works well for many operations, but meeting coordination is often not a single call. It can require context, clarification, negotiation, and several turns before the task is complete.
That is one reason we believe specialized agents will matter. Some workflows are better represented as a capability owned by another agent than as a collection of low-level tools.
Meeting coordination is one of those workflows.
What is live today
Ask My Envoy now exposes a production MCP server, is published in the official MCP Registry, and has a working A2A interface. Those interfaces sit on top of the same coordination platform used by Envoys today.
The protocol surface is still evolving, and we expect the implementation to change as the agent ecosystem matures. The underlying direction is more stable: when a generalist AI agent needs to coordinate a meeting, it should be able to work with an Envoy rather than rebuild meeting coordination itself.
Building for an agent-to-agent web
The web moved from documents to applications, then from applications to APIs. AI agents add another layer, where software does not just expose functions but increasingly exposes capabilities that other agents can discover and use.
Ask My Envoy is being built for that environment. The role of the Envoy is not to become another generic assistant, but to become the specialized agent that other agents can rely on when the task is meeting coordination.
That is the technical direction behind our work with MCP, A2A, agent discovery, Agent Cards, and cross-organization scheduling.