llmconveyors-mcp

Guidelines for using LLM Conveyors MCP tools effectively

# LLM Conveyors MCP Usage Rules

## API Key
- The LLMC_API_KEY environment variable must be set with a key starting with `llmc_`
- Different tools require different API key scopes. If you get a scope error, tell the user which scope is needed.

## Agent runs
- `job-hunter-run` and `b2b-sales-run` take 1-10 minutes to complete. Let the user know it will take a moment.
- These tools consume credits. Always confirm with the user before running an agent.
- Agent runs return a `sessionId` and `generationId` that can be used with `session-hydrate`, `share-create`, and `document-download`.

## Resume format
- `resume-validate`, `resume-render`, and `resume-preview` expect JSON Resume format (an object with `basics`, `work`, `education`, `skills` sections).
- `ats-score` expects plain text strings, not JSON objects.
- Master resumes use `label` (name) and `rawText` (plain text content), not JSON Resume format.

## File uploads
- `upload-resume` and `upload-job-file` accept base64-encoded file content. Encode the file bytes to base64 before calling.
- `upload-job-text` accepts plain text directly, no encoding needed.

## Sessions
- `session-create` takes no required parameters. Metadata is optional.
- `session-hydrate` returns the full session with all artifacts and logs. Use this to retrieve generation results.
- Always clean up test sessions with `session-delete` when done.

## Content and sharing
- `content-save` requires `docType` (one of: original_cv, extensive_cv, cover_letter, cv_strategy, cover_letter_strategy, cold_email_strategy, recon_strategy, company_context) and `content` (string).
- `content-delete-generation` requires both the generation `id` and the `sessionId`.
- `share-create` requires `sessionId` and `generationId`. Share links are rate-limited to 1 per week.

MCP server: name

"llmconveyors-mcp"

MCP server: description

"39 tools for the LLM Conveyors AI agent platform. Run Job Hunter, B2B Sales, ATS scoring, resume rendering, and more from any MCP client."

MCP server: vendor

"LLM Conveyors"

MCP server: homepage

"https://llmconveyors.com"

MCP server: sourceUrl

"https://github.com/ebenezer-isaac/llmconveyors-mcp"

MCP server: license

"MIT"

MCP server: runtime

"node"

MCP server: command

"npx"

MCP server: args

[
  "-y",
  "llmconveyors-mcp"
]

MCP server: env

{
  "LLMC_API_KEY": {
    "description": "LLM Conveyors API key (starts with llmc_)",
    "required": true
  }
}

Score a resume against a job description for ATS (Applicant Tracking System) compatibility. Use when the user wants to check how well their resume matches a job posting or improve their resume's keyword coverage.

# ATS Score

Score a resume against a job description using keyword matching and AI semantic analysis.

## When to use

- User wants to know how well their resume matches a job posting
- User asks about ATS compatibility or keyword optimization
- User wants to improve their resume for a specific role
- After generating a tailored CV with Job Hunter, to validate the output

## Steps

1. **Get the resume text.** Either ask the user to paste their resume, or use `upload-resume` to parse a file, or use a master resume from `master-resume-list`.

2. **Get the job description.** Ask the user for the full job description text, or use `upload-job-text` to parse it.

3. **Score it.** Call `ats-score` with:
   - `resumeText` - the resume as plain text
   - `jobDescription` - the job description as plain text

4. **Present the results.** The response includes:
   - `overallScore` - numeric score (0-100)
   - `grade` - letter grade (A through F)
   - `breakdown` - scores for keyword match, experience relevance, skills coverage, education fit, format quality
   - `matchedKeywords` - which keywords were found
   - `missingKeywords` - which keywords are missing
   - `suggestions` - specific improvements to make

5. **Help improve.** Walk through the missing keywords and suggestions with the user.

## Example

User: "How does my resume score against this job posting?"

```
1. Get resume text and job description from the user
2. Call ats-score with both as plain text strings
3. Present the score, grade, and specific improvement suggestions
```

Run the B2B Sales agent to research a company and generate personalized sales outreach. Use when the user wants to prospect a company, write cold emails for sales, or research a target account.

# B2B Sales

Research a target company and generate personalized cold email outreach for B2B sales.

## When to use

- User wants to research a company for sales prospecting
- User needs a personalized cold email for a specific company
- User mentions B2B outreach, sales emails, or account research
- User has a company name and website they want to target

## Steps

1. **Get company details.** You need: company name and website URL. Ask the user if not provided.

2. **Run the agent.** Call `b2b-sales-run` with:
   - `companyName` - target company name
   - `companyWebsite` - company website URL
   - `strategy` (optional) - specific angle or approach for the outreach

3. **Present the results.** The agent returns company research, person research (identifying the right contact), and a drafted cold email. Share these with the user.

4. **Create a share link.** If the user wants to share the output, call `share-create` with the session and generation IDs from the run result.

## Example

User: "Research Notion and draft a sales email for our developer tools product"

```
1. Call b2b-sales-run with companyName="Notion", companyWebsite="https://notion.so", strategy="Focus on developer tools integration"
2. Wait for completion (may take 3-10 minutes)
3. Present the company research report and drafted cold email
```

## Tips

- The agent performs deep company research including identifying decision makers
- Results typically take 3-10 minutes depending on how much public information is available
- The strategy parameter helps focus the outreach angle

Run the Job Hunter agent to generate tailored CVs, cover letters, and cold emails for job applications. Use when the user wants to apply for a job, tailor their resume, or write outreach emails for a specific role.

# Job Hunter

Generate a complete, tailored job application package for a specific role.

## When to use

- User wants to apply for a job and needs a tailored CV
- User needs a cover letter written for a specific position
- User wants cold outreach emails to hiring managers or recruiters
- User has a job description and wants their resume optimized for it

## Steps

1. **Get the job details.** You need at minimum: company name, job title, and job description. Ask the user if they haven't provided these.

2. **Check for a master resume.** Call `master-resume-list` to see if the user has saved master resumes. If they have one, use its ID with the `masterResumeId` parameter.

3. **Run the agent.** Call `job-hunter-run` with the job details:
   - `companyName` - target company
   - `jobTitle` - the role
   - `jobDescription` - full job description text
   - `theme` - resume theme (default: "even", options: even, stackoverflow, class, professional)
   - `mode` - "standard" for CV + cover letter, "cold_outreach" to also generate cold emails
   - `contactName` and `contactEmail` - if the user has a specific hiring contact

4. **Present the results.** The agent returns artifacts including the tailored CV, cover letter, and optionally cold emails. Share the key outputs with the user.

5. **Score it.** Optionally call `ats-score` with the generated resume text and the job description to show the user how well their tailored CV matches.

## Example

User: "I want to apply for the Senior Engineer role at Stripe. Here's the job description: ..."

```
1. Call job-hunter-run with companyName="Stripe", jobTitle="Senior Engineer", jobDescription="..."
2. Wait for completion (may take 2-3 minutes)
3. Present the tailored CV, cover letter, and ATS score
```

## Tips

- The agent takes 1-3 minutes to complete as it researches the company and generates multiple documents
- If the user has uploaded their resume before, it will be used automatically
- Cold outreach mode requires a contact name and email for best results

Validate, render, preview, and manage resumes. Use when the user wants to check a resume for errors, generate a PDF, preview as HTML, list available themes, or manage their saved master resumes.

# Resume Tools

Full suite of resume management tools including validation, rendering, and master resume CRUD.

## When to use

- User wants to validate a resume for errors or warnings
- User wants to render a resume as PDF or preview as HTML
- User asks about available resume themes
- User wants to save, list, update, or delete master resumes
- User wants to import from or export to Reactive Resume format

## Available tools

### Validation and rendering
- `resume-validate` - Check a JSON Resume for errors and warnings
- `resume-render` - Generate a PDF from a JSON Resume (returns base64 PDF)
- `resume-preview` - Generate an HTML preview of a resume
- `resume-themes` - List available themes (even, stackoverflow, class, professional)

### Master resume management
- `master-resume-create` - Save a new master resume (needs `label` and `rawText`)
- `master-resume-list` - List all saved master resumes
- `master-resume-get` - Get a specific master resume by ID
- `master-resume-update` - Update a master resume's label or text
- `master-resume-delete` - Delete a master resume

### Import/Export
- `resume-import-rx` - Convert Reactive Resume format to JSON Resume
- `resume-export-rx` - Convert JSON Resume to Reactive Resume format

## Steps for common tasks

### Save a master resume
1. Ask the user for their resume text
2. Call `master-resume-create` with a `label` (name) and `rawText` (the resume content)
3. Confirm the save and share the ID for future reference

### Render a PDF
1. Get or build a JSON Resume object (with basics, work, education, skills sections)
2. Optionally call `resume-themes` to let the user pick a theme
3. Call `resume-render` with the resume object and chosen theme
4. The response contains a base64-encoded PDF

Upload and parse resumes and job descriptions from files or text. Use when the user provides a resume file, job posting URL, or job description text that needs to be parsed into structured data.

# Upload and Parse

Parse resumes and job descriptions into structured data for use with other tools.

## When to use

- User provides a resume file (PDF, DOCX, TXT) to upload
- User pastes a job description or provides a job posting
- User wants to extract structured data from a document before scoring or generating

## Available tools

- `upload-resume` - Upload and parse a resume file. Accepts base64-encoded content. Returns structured resume data.
- `upload-job-file` - Upload and parse a job description file. Accepts base64-encoded content.
- `upload-job-text` - Parse a job description from plain text. Returns company name, job title, keywords, and other structured fields.

## Steps

### Parse a job description from text
1. Call `upload-job-text` with the `text` parameter containing the job description
2. The response includes parsed fields: `companyName`, `jobTitle`, `keywords`, `metadata`
3. Use these fields with other tools like `ats-score` or `job-hunter-run`

### Upload a resume file
1. The file must be base64-encoded. If working with a local file, read it and encode to base64.
2. Call `upload-resume` with `fileBase64` (the encoded content), `filename` (original name), and optionally `contentType`
3. The response includes the parsed and normalized resume data

### Upload a job description file
1. Same process as resume: base64-encode the file
2. Call `upload-job-file` with `fileBase64`, `filename`, and optionally `contentType`
3. Returns the same structured output as `upload-job-text`