fps-cv-mcp
The server fetches CV data from GitHub Release assets and uses GitHub Actions for publishing and deploying updates.
GitHub Actions automates publishing and deploying the CV when data is updated.
The CV includes a Google Scholar profile link that can be queried.
The MCP server is hosted on Wasmer Edge, and CV updates are deployed to Wasmer apps.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@fps-cv-mcp Summarize Francisco's CV for a technical hiring manager"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
CV Agent Toolkit
A suite of tools for working with Francisco Perez-Sorrosal's CV and professional information. Combines an MCP server, two agent skills, and two Claude Code plugins into installable packages.
What It Does
MCP Server — serves CV content in markdown, PDF, LaTeX, HTML, and Typst formats via 16 tools, plus semantic query capabilities and tailored CV rendering
cv-analystSkill — structured CV summarization for different audiences (hiring screens, executive briefings, technical reviews) with output in multiple formatscv-tailoringSkill — job-targeted CV tailoring that analyzes a job description, selects relevant content, and produces a page-constrained (2-3 pages) compiled PDFTwo Claude Code Plugins —
cv(consumer-facing) andcv-forge(maintainer-facing, editor + publisher)
Related MCP server: Interactive Curriculum MCP Server
Installation
Claude Code (Recommended)
First-time install
Install the consumer plugin from the marketplace:
claude plugin marketplace add francisco-perez-sorrosal/bit-agora
claude plugin install cv@bit-agora --scope userBoth the cv (consumer) and cv-forge (maintainer) plugins are available. Install cv for end-user CV access; additionally install cv-forge if you need to edit CV data or manage releases.
Configuration (for the maintainer plugin only)
The maintainer plugin needs the paths of your local clones. Pass them at install time; the values are validated against the manifest and stored in your user settings:
claude plugin install cv-forge@bit-agora --scope user \
--config cv_repo_path=/Users/you/dev/cv \
--config cv_forge_path=/Users/you/dev/cv-forgeBoth paths must be absolute. cv_repo_path is required; cv_forge_path is optional (without it the skills fall back to a cv-forge executable on PATH, and the deploy skill refuses to run). To change a value, uninstall and reinstall with the new --config. The interactive /plugin configure command only sees project-scoped installs, so it reports the user-scoped plugin as "not installed in this project".
Update plugins
To pick up new releases of the plugins:
claude plugin update cv@bit-agora
claude plugin update cv-forge@bit-agoraThen restart Claude Code.
Claude Desktop
The server is remote, so there is nothing to download or run locally. Add it as a custom connector:
Open Claude Desktop → Settings → Connectors
Choose "Add custom connector"
Name it (for example
Francisco's CV) and enter the URLhttps://fps-cv-mcp.wasmer.app/mcp; no authentication is requiredSave, then enable the connector from the tools menu of a conversation
The server always serves the latest published release of the cv repository; you never need to update the connector after a new CV release. The raw tools (get_cv, get_cv_sections, query_work, …) and summarize_cv are available as soon as the connector is enabled.
Skills in Claude Desktop and claude.ai
The cv-analyst and cv-tailoring skills ship inside the Claude Code plugin; Claude Desktop and claude.ai take them as uploaded zips instead. Build them from this repository:
make build-skillThis writes dist/skill/cv-analyst.zip and dist/skill/cv-tailoring.zip (each contains SKILL.md and its references/; the frontmatter is limited to the portable name and description fields the upload accepts). Then, in Claude Desktop or claude.ai: Settings → Features (or Capabilities) → Skills → Add skill → upload one zip at a time. Rebuild and re-upload after a plugin release that changes a skill.
Without the skills, summarization still works through the summarize_cv tool; tailoring (job-targeted CV with compiled PDF) needs cv-tailoring plus a local LaTeX or Typst toolchain, so it is best used from Claude Code.
Usage
Retrieve and Summarize the CV
"Get Francisco's CV"
"Summarize Francisco's CV for a startup executive briefing"
"What is Francisco's Google Scholar profile link?"
"Give me a 3 page summary of my CV for a hiring manager oriented towards an AI engineer position in HTML"
"Give me the CV in LaTeX"
Tailor the CV for a Job
Use the cv-tailoring skill to adapt the CV to a specific job description:
"Tailor my CV for this job: [paste job description]"
"Adapt my resume for a Senior ML Engineer position at Google"
"Customize my CV for this LinkedIn job"
The tailoring pipeline analyzes the job description, assesses fit against CV content, reorders sections, filters entries, and renders a 2-3 page PDF.
Publish (Maintainer Only)
Use the cv-forge plugin to edit CV data, open a PR, publish a release, and deploy updates:
"Edit my CV: change my title at Yahoo to Principal Research Engineer"
"Show me the diff and open a PR" (validation happens automatically)
"Publish the CV" (tags and deploys to Wasmer)
The Two Repositories
cv (data repository) — Contains the CV content in structured YAML format:
cv-data/resume.yaml— Work experience, education, projects, publications, skillscv-data/resume-semantics.yaml— Semantic overlay with topic taxonomy and cross-referencesschemas/— JSON Schema pair for validation and structure definition
cv-forge (this repository) — Contains all the machinery:
Python MCP server and CLI for rendering and serving
Jinja2 templates for five output formats
Two Claude Code plugins
GitHub Actions for publishing and deploying
Wasmer Edge apps for hosting
The two repositories share no code — only two pinned artifacts (a reusable GitHub workflow and mirrored schemas) and one runtime data source (GitHub Release assets).
Rendered CV
The latest compiled CV is published as:
HTML (interactive) — https://fps-cv.wasmer.app/
PDF — GitHub Release asset at
releases/latest/download/resume.pdfLaTeX (moderncv) — GitHub Release asset at
releases/latest/download/resume.texTypst (moderner-cv) — GitHub Release asset at
releases/latest/download/resume.typstMarkdown — GitHub Release asset at
releases/latest/download/resume.md
All assets are regenerated every time the CV data is updated.
Release Assets
Each release contains eight stable, version-free assets:
Asset | Format | Purpose |
| Markdown | Full CV for AI consumption |
| LaTeX (moderncv) | Full CV for local compilation |
| HTML | Interactive CV for web browsers |
| Typst (moderner-cv) | Full CV for Typst compilation |
| Compiled full CV | |
| LaTeX (tailored) | Tailored template (requires TailoringSpec) |
| Typst (tailored) | Tailored template (requires TailoringSpec) |
| JSON | Release metadata and asset manifest |
For developer documentation and deployment details, see README_DEV.md and RELEASE_PROCESS.md.
For the CV data repository, see github.com/francisco-perez-sorrosal/cv.
Support
For technical issues or questions about the machinery, refer to this repository. For questions or corrections about CV content, refer to the cv data repository.
Available Tools
16 toolsget_cvARead-onlyIdempotent
Data-layer tool: retrieves raw CV content in markdown, PDF, LaTeX, HTML, or Typst source.
When the cv-analyst skill is available, prefer invoking that skill instead of calling this tool directly — the skill orchestrates retrieval with proper formatting, artifact delivery, and summarization.
format='markdown' (default): LLM-readable text for analysis. format='pdf': original PDF binary for inline rendering. format='latex': full CV as LaTeX source (moderncv package) for typeset PDF generation. format='html': self-contained interactive HTML with theme switching and expandable cards. format='typst': full CV as Typst source (moderner-cv package) for typeset PDF generation.
Inputs: enrich, format.
get_cv_pdf_linkARead-onlyIdempotent
Return the direct link to the PDF version of the CV.
Inputs: no named parameters. Returns: result
get_cv_sectionsARead-onlyIdempotent
Retrieve one or more CV sections in a single call.
Accepts a list of section names. Returns all matched sections separated by blank lines. Reports any unrecognized names with the list of available sections.
Inputs: section_names (required), enrich. Returns: result
get_entryARead-onlyIdempotent
Retrieve a specific resume entry by its stable ID. Returns JSON representation.
Inputs: entry_id (required). Returns: result
get_entry_contextARead-onlyIdempotent
Get full semantic context for an entry: topics, relationships, summaries, impact.
Inputs: entry_id (required). Returns: result
get_google_scholar_linkARead-onlyIdempotent
Return the Google Scholar profile link.
Inputs: no named parameters. Returns: result
get_linkARead-onlyIdempotent
Return a profile or document link by network name.
Inputs: name (required). Returns: result
get_relationshipsBRead-onlyIdempotent
Get cross-references and relationships for a resume entry.
Inputs: entry_id (required). Returns: result
get_skill_profileARead-onlyIdempotent
Get skill proficiency levels across the career, optionally filtered by topic.
Inputs: topic. Returns: result
get_tailored_cvARead-onlyIdempotent
Render a tailored CV from a TailoringSpec in LaTeX or Typst format.
The tailoring config controls section ordering, entry emphasis, and profile override. Returns compilable source code. Use this tool after analyzing a job description to produce a targeted CV.
Inputs: tailoring_config (required), format. Returns: result
list_cv_sectionsARead-onlyIdempotent
List available CV section names with approximate line counts.
Helps AI assistants pick the right section for targeted queries.
Inputs: no named parameters. Returns: result
list_entry_idsARead-onlyIdempotent
List all entry IDs with labels, optionally filtered by section type (work, patents, publications, education, certificates, conferences, memberships, skills, book_reviews).
Inputs: section. Returns: result
list_linksARead-onlyIdempotent
List all available profile and document links.
Inputs: no named parameters. Returns: result
query_by_topicBRead-onlyIdempotent
Find resume entries annotated with a topic. Returns entry IDs with labels.
Inputs: topic (required), include_subtopics. Returns: result
query_workBRead-onlyIdempotent
Filter work entries by company, date range, or topic. Returns matching entries as markdown.
Inputs: topic, enrich, company, end_year, +1 more. Returns: result
summarize_cvARead-onlyIdempotent
Fallback CV summarization for clients without Agent Skills support.
When the cv-analyst skill is available, prefer invoking that skill instead — it provides richer orchestration, preset profiles, and artifact delivery. This tool exists for MCP clients that cannot load skills.
Inputs: tone, style, context, depth_level, +6 more. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 16 tools
Most tools are distinct (list vs. query vs. get vs. summarize), but several overlap: query_by_topic and query_work both search by topic, and get_cv_pdf_link/get_google_scholar_link are convenience wrappers around get_link while get_cv can also return PDF. Descriptions help clarify the boundaries, but an agent could still misselect between related retrieval and link tools.
All tool names use a consistent snake_case verb_noun or verb_object pattern: list_*, get_*, query_*, summarize_cv. The naming is highly predictable and makes it easy to guess the general purpose of a tool before reading the description.
At 16 tools is just past the typical 3-15 well-scoped range, and the set includes redundant link/query variants such as get_cv_pdf_link, get_google_scholar_link, get_link, query_by_topic, and query_work. It is not extreme, but the surface feels slightly heavy for a read-only CV/resume server.
For a read-only CV retrieval and tailoring domain, the coverage is strong: raw CV formats, sections, individual entries, semantic context, skill profiles, relationships, links, and tailored CV rendering are all available. Minor gaps exist only in trimming redundancy rather than missing core functionality.
Maintenance
Related MCP Connectors
Your CV as a document you own, edited by your AI assistant, in several languages, as a typeset PDF
Build, version and render resumes as PDFs from Claude or any MCP client.
Tailor resumes, generate cover letters, render CVs as PDF, and browse 22+ templates.
Store your CV once, then get tailored .docx resumes, cover letters and keyword gaps for a job.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceExposes a structured professional resume as a set of AI-queryable tools, enabling AI clients like Claude Desktop to query summary, experience, skills, projects, and tailor resumes to job descriptions.1MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI assistants and LLM clients to query a professional CV and portfolio, including work history, technical skills, projects, job compatibility evaluation, education, and contact details.MIT
- AlicenseNot gradedqualityCmaintenanceEnables drafting, validating, and reviewing tailored CVs and cover letters from private local career inputs, with reproducible PDF builds and application tracking support.MIT
- AlicenseAqualityCmaintenanceExposes a curriculum vitae hosted on a backend REST API to AI assistants over stdio, letting them retrieve the full CV, list and fetch individual sections, and search for terms like technologies or companies. It also provides the full CV as attachable context and a prompt that compares the CV against a job description to suggest improvements.4MIT