What Is an MCP Prompt Template?

An MCP prompt template is an instruction set that tells an AI agent what task to perform and how to perform it. Think of it as the difference between handing someone a hammer (the tool) and saying "build a deck by next Friday" (the prompt). The template is the "what do I do?" layer—it contains the goals, constraints, and decision rules the agent follows when executing a job search or application workflow.

In MCP (Model Context Protocol) systems, a prompt template includes things like:

  • What kind of jobs to search for (seniority level, location, salary range)
  • Which job boards to query
  • How to filter out irrelevant postings (ghost jobs, low-match positions)
  • Whether to auto-apply, cold-outreach, or just log results
  • How to handle rejections or application failures

You can change the prompt without touching the underlying tool. You reuse the same search tool, but the prompt template changes its behavior—from "find me full-time W-2 roles" to "find me first-to-apply C2C contracts" in moments.

How a Prompt Template Differs From a Tool

A tool is the executable code—the API caller, the parser, the database connector. It's the infrastructure. A prompt template is the choreography that tells the tool when and how to work.

Example: A job-board search tool can hit Indeed, LinkedIn, or Dice. The tool itself doesn't care. The prompt template decides whether you search Indeed first, which filters to apply, and whether to backfill from LinkedIn if the first board has no matches.

You can have one tool and hundreds of prompt templates. Different templates might be designed for:

  • Software engineers hunting remote roles
  • C2C contractors looking for first-to-apply eligible postings
  • Passive candidates who want cold outreach only, no auto-apply
  • High-velocity searchers who apply to everything above a minimum match score

The tool stays the same. The prompt template is what makes the agent behave differently for each use case.

How a Prompt Template Differs From a Resource

A resource is the data or service the tool accesses. Examples: a job board's database, a resume parser, a rate-limit quota, or a list of recruiter contact info.

The prompt template decides which resources to use and in what order. It's the instruction that says "check the real-time feed first, then fall back to the static board if you need more results." The feed and the board are resources. The decision logic is the template.

Think of it this way:

  • Tool: The ability to call an API
  • Resource: The API endpoint and the data behind it
  • Prompt template: The logic that says "call this API, parse this field, filter by these rules, then call the next API if needed"

Why This Distinction Matters for Job Agents

If you're using an AI job agent like GiraffyReach, understanding these primitives helps you troubleshoot and optimize. When an agent isn't finding the jobs you want, the fix might be:

  • A better prompt template (change the search strategy)
  • A resource problem (the job board is slow or the API quota is exhausted)
  • A tool issue (the parser is breaking on a new HTML structure)

Each layer requires different fixes. A prompt template tweak is fast and free. A tool upgrade or resource expansion might require engineering time. Knowing which layer is broken saves weeks of debugging.

Practical Example: Two Templates, One Tool

Imagine a job-search tool that queries LinkedIn, Indeed, and a proprietary C2C board. One prompt template might say:

"Search all three boards for 'DevOps Engineer' roles in the US, filter for remote only, rank by recency, auto-apply to the first 5 matches per day, skip if salary below $120k, pause if rejection rate exceeds 20%."

Another template using the same tool might say:

"Query only the C2C board, filter for first-to-apply eligible contracts, rank by rate (highest first), do cold outreach to the recruiter instead of auto-applying, include salary data in each email, limit to 3 outreach per week."

Same tool. Different instructions. Different outcome.

Understanding this three-layer model—prompt, tool, resource—is essential for anyone serious about job-search automation. It lets you see past the marketing and understand what you're actually controlling and what you're not.