AuthorAgent — Cover Letter
Score and build cover letters against a 100-point rubric — powered by Claude AI.
A free tool that evaluates your cover letter across 10 categories and 100 points — then helps you build a better one. Runs inside your Claude account. Takes 2 minutes to set up.
Get Started — Free Setup Guide ↓What It Does
Paste any cover letter and get a visual scoring dashboard. The agent evaluates 10 categories — from whether your opening sentence makes a claim about the work to whether your closing avoids repeating evidence from the body. Every deduction comes with a specific quote from your letter and a concrete fix.
Upload your resume and a job description. The agent researches the company, asks you a few targeted questions, presents its strategy for your approval, then generates a 675–700 word cover letter designed to score 90+ on the rubric. You see the letter and its score together.
Score an existing letter first. See exactly where it loses points. Then let the agent rewrite it targeting the specific deductions — and score the new version so you can compare.
How It Works
Go to claude.ai and sign up. The free plan works. A paid plan (Pro, $20/month) gives you more messages and access to stronger models.
Click "Projects" in the left sidebar. Click "Create Project." Give it any name you like — "Cover Letter Agent" works fine.
You will find a field called "Custom Instructions" (sometimes labeled "Project Instructions"). This is where the agent gets its brain. Click the button below to copy the instructions, then paste them into that field.
Click "New Chat" inside your project. Type "hi" or anything at all. The agent will immediately present three buttons: Score a letter, Build a letter, or Score then rebuild. Pick one and follow the prompts.
That's it. No coding. No API keys. No installation.
What You Need Ready
Here is what to have ready depending on what you want to do:
The Rubric
Your letter is evaluated on a 100-point scale across these categories:
Word count, paragraph count, no em-dashes, no single-sentence paragraphs. These are hard rules with no partial credit.
Does one governing argument control the entire letter? Can every sentence justify its existence? Does the space given to each topic match what the job description emphasizes?
Does your first sentence make a claim about the nature of the work (not about you)? Do you enter as proof in sentence two? Does every paragraph after the first open with a capability claim rather than scene-setting?
Does the letter explicitly answer three questions: why this company, why now, and why you are leaving your current path? All three must be present.
The single largest category. The agent identifies 4–5 primary responsibilities from the job description and checks whether your letter covers each with a named framework, a specific metric, and structural detail proving you owned the work.
Do your metrics explain how you work, not just what you delivered? Is there a business performance metric proving commercial viability? Does every evidence paragraph connect to what the target company specifically gains?
Does the letter use exact names from the company's own materials? Are at least two company-specific details woven into the argument so that removing them would weaken it?
No permission framing ("given the opportunity to lead"). No banned phrases ("look forward to discussing," "passionate about," "strong track record"). Active voice. Executive register.
Is the reason you want this specific job anchored to something only this company offers? Is it framed as what the company gains, not what you need?
Does the closing frame your capabilities as a proven pattern, not a promise? Does the final sentence name the company? Does the closing avoid restating evidence from the body?
Model Guide
The agent works on any Claude model, but output quality varies.
| Plan | Messages | Expected Score | Best for |
|---|---|---|---|
| Free — Sonnet | ~15–30 per 5 hrs | 75–85 | Trying the tool, scoring existing letters, one-off applications |
| Pro — Sonnet Recommended | ~45 per 5 hrs | 78–88 | Regular job applications, iterating on multiple versions |
| Pro — Opus | ~10–15 per 5 hrs | 85–92 | High-stakes applications where letter quality matters most |
| Max — Sonnet or Opus | 5× or 20× Pro capacity | 85–92 | Active job searches, multiple roles at once |
Use Sonnet. It handles the full workflow reliably, gives you more sessions per window, and produces strong letters. Switch to Opus only for your top-priority applications.
FAQ
The agent instructions are completely free. You need a Claude account to use them — Claude offers a free plan, or you can subscribe to Pro ($20/month) or Max ($100–200/month) for more capacity.
No. You copy text into a field and start chatting. There is no coding, no installation, and no API keys involved.
The agent is designed for senior professional roles — manager, lead, director, VP level — but works for any role where a cover letter benefits from structured argumentation and specific evidence. The rubric adapts to any industry; the agent reads each job description and calibrates its scoring categories to match.
Yes. Start a new chat inside the same project for each application. The agent instructions stay loaded; you just provide a new resume/JD combination each time.
No. Claude does not retain conversation data between sessions (unless you have memory features enabled in your account settings). Your resume, letters, and job descriptions are only available during the active chat.
Scoring a letter: 3–5 minutes. Building a letter from scratch: 10–15 minutes. Scoring and rebuilding: 15–20 minutes. These times assume you have your resume and JD ready to paste.
The questions fill gaps that your resume cannot answer: why you are leaving your current path, what your relationship with this specific company is, and what the largest financial decision you personally owned was. These details are what separate a generic cover letter from one that scores 90+.
The scoring is based on a specific rubric, not subjective judgment. Every deduction comes with a quote from your letter and an explanation of what the rubric requires. You can ask the agent to "dive deeper" on any category to see the full reasoning.
About
This tool was built using over 140 letter generations and 6 agent model iterations prior to deployment. The rubric, the eight pre-writing decisions, the five-paragraph construction sequence, the seven failure modes, and the eight-point audit checklist all encode specific lessons from rounds of generation → scoring → diagnosis → fix.
The system was calibrated against a 100-point rubric across 10 categories and validated through independent scoring.