Card 05 · AI agents, explained
What is MCP? The plug, not the appliance
The Model Context Protocol is the open standard that lets an AI app call outside tools, so a calendar, a database or a CRM is wired once and works in every AI that speaks it. MCP is the plug; the tool on the other end is what does the work.
01 — The problem
Every tool used to need its own build.
Calendar, accounting, CRM: each arrived as a custom integration, priced and kept alive separately. Adding the fourth tool meant another small project.
02 — The standard
MCP is one plug both sides agree on.
The Model Context Protocol is a published standard for how an AI app asks a tool what it can do and then calls it. The same plug sits on both ends, so neither side builds for the other.
03 — The tool's side
The tool publishes a server.
Whoever owns the calendar publishes a small server that answers what it can do, in the shapes the protocol defines. One server serves every AI that speaks MCP.
04 — Your side
Your agent is the client.
The agent asks the server what tools exist, then calls one by name with arguments. Nothing in the model needs to know what a calendar is in advance.
05 — What travels
Names, arguments and results.
That is the whole payload, which is why a new tool needs no bespoke prompt written for it. The tool still does the work; the protocol only carries the request.
06 — What changes
A fourth tool becomes a connection.
In TalentOS the same harness reaches accounting, chat and calendar without a build for each. The fourth tool is an afternoon, not a quarter.
07 — What it is not
It is not a security model.
The protocol says what a tool can do, not who may call it. What an agent may run on its own is a separate setting, per action and per person.
08 — Start here
Connect the one tool the team already uses.
Read access to the calendar or the shared drive is enough to see whether the loop works. Add the second one with write access only after you have watched a week of runs.
Each server is published by whoever owns the tool.
One plug, three tools today, and the fourth arrives as a connection rather than a project.
Tools for this
Every price and licence below is what the tool’s own page said on 18 November 2026, with two sources behind each record. There are no scores and no order of preference; where TalentOS is one of the tools it is marked ours and listed like the rest.
A connector platform that gives agents authenticated access to hundreds of apps, with tool calling, per-tool permissions and firewall policies.
For a team that needs its agent to act in Gmail, Slack, HubSpot or Stripe without building each integration and its sign-in.
$18 per seat per month · Closed source
The official index of MCP servers, so a client or a person can find a server by name instead of hunting for a URL.
For a developer publishing a server, or a team checking whether an integration already exists before building one.
$0 (free) · Open source, Apache-2.0
An open standard that lets an AI app discover and call outside tools, so one server works in every client that speaks it.
For a developer who wants to write an integration once, and a team deciding what its agent is allowed to reach.
$0 (free) · Open source, MIT
Model Context Protocol documentationSpecification repository and licence
A developer platform where each app connector also works as an MCP tool, with managed authentication for thousands of APIs.
For a developer who wants production connectors and MCP endpoints without maintaining the sign-in for every service.
$0 to build, paid from $99 per month · Closed source
An automation platform whose MCP server exposes app actions to an AI client, running them on the app connections a team already has in Zapier.
For a team already paying for Zapier that wants its agent to act in the apps it has connected, and its recurring work on a schedule.
$19.99 per month (usage) · Closed source
A shared workspace where each person brings their own AI agent into Rooms with the team, with company memory, skills, routines, permissions and a record of what each agent did.
For a team of about 5 to 50 people that wants its people and their agents working in one place, with each answer and each approval attributed to a person.
$20 per member per month (seat) · Closed source