Most AI tools let you choose which model you are talking to, and with TaskRay MCP that choice matters more than it does in regular chat. This article explains which model to reach for, and how the effort setting changes what you get.
In this article:
- Why the model choice matters more with MCP
- Picking the right tier
- The effort dial
- A note for Copilot and Agentforce users
Why the model choice matters more with MCP
MCP work is mostly tool calls: look up a project, list some tasks, create a record. Smaller, faster models handle that kind of work beautifully while responding much more quickly. For a quick answer or a status check, a fast-tier model with a low effort setting gets you there sooner, with no loss in quality on the task at hand.
Picking the right tier
Match the model to the job. Most of the time the daily driver is all you need, and you step up only for genuinely hard analysis.
| Tier | Claude | ChatGPT | Use it for |
|---|---|---|---|
| Fast | Haiku | Luna | Quick answers and status checks |
| Daily driver (recommended) | Sonnet | Terra | Complex analysis, portfolio health, cross-project planning |
| Synthesis | Opus | Sol | Your toughest challenges |
The effort dial
Effort is the other setting. Most tools let you choose how hard the model thinks before it answers. Higher effort helps on hard problems, and it just adds waiting on simple ones. For most prompts in this guide, low effort is all you need. Running the portfolio and leadership prompts on a daily-driver model (Sonnet or Terra) at medium effort is the ceiling most teams will ever reach.
A note for Copilot and Agentforce users
If you work in Microsoft Copilot or Agentforce, your platform or admin usually manages model choice for you, so this article is mainly for Claude and ChatGPT users. The prompts themselves work the same way everywhere.
Where to go next
- Getting Started with TaskRay MCP, for how access and approvals work.
- First Prompts to Try After Connecting TaskRay MCP, to put a model to work.