Why we never meter MCP

Every project tool now has an agent story, and many of them come with a meter attached: agent credits, AI plans, calls that count against an allowance, features that start at the second-highest tier.

NEWWORK Projects doesn’t meter MCP. On every plan, including Free, your agents can read briefs, start tickets, tick checks, link pull requests and run timers as often as they need. We don’t count the calls.

The work is the agent’s, the decisions are yours

A coding agent working through a sprint makes hundreds of small calls: what’s assigned to me, what does this ticket say, I’m starting, this check passed with this pull request, I’m stopping the timer. Each of those calls is the project staying true. A meter on them is an incentive to call less, which means a project that knows less about what happened.

We would rather you connect every agent you use and let them keep the tickets honest.

What we do charge for

We charge per member, for the people who decide what gets built: the plans, goals and reports that need a team to steer them, and the business side of the work, such as capacity, budgets and change requests. Guests, such as your clients, are free.

An agent connects with a member’s personal token and acts as that person, so it is never a seat of its own. Ten agents working for one engineer cost the same as that engineer.

The one thing that is limited

The built-in AI that writes briefs, Pulses and brainstorms runs on our OpenAI account, so each plan includes a monthly allowance of AI actions. Add your own OpenAI key under Integrations and the limit is gone, on every plan.

That is the whole model. See pricing for the numbers.