Skip to main content
Glama

Recruiter Agent

An AI agent that answers recruiters' questions about the skills, projects and experience of a software engineer. The engineer is an openly fictional persona. This repository is a portfolio project showing how to build, test and ship a remote MCP server on AWS.

Status: early setup. Sections marked TBD are filled in as the project progresses.

What it is

  • A remote Model Context Protocol server (TypeScript, streamable HTTP, stateless) running on AWS Lambda behind a Function URL.

  • The persona profile is stored as files in S3.

  • Any MCP client (Claude, Cursor, ChatGPT developer mode) can connect and use its own model.

  • If the profile has no answer, the tools return an explicit "not found" result. The agent never invents skills.

Related MCP server: Portfolio MCP

Architecture

TBD: C4 diagrams in docs/architecture.md.

Key decisions are recorded as ADRs in docs/adr/. Day-to-day decisions and lessons learned go in docs/decision-log.md.

Connect an MCP client

TBD: Claude custom connector, Claude Desktop, Cursor.

Development

Prerequisites: Node.js 22+, gitleaks (required by the pre-commit hook), Terraform (for infra/).

npm install        # also activates git hooks (.githooks/)
npm run lint       # ESLint + Prettier check
npm test           # Vitest unit tests
npm run build      # compile to dist/

Repository layout:

Path

Contents

src/

MCP server source

tests/

Unit tests (tool contracts)

evals/

Golden question set and eval runner

infra/

Terraform

specs/

Persona and feature specs

docs/

ADRs, decision log, architecture

Built with Claude Code

TBD. The guardrails are committed in .claude/settings.json: secret files are blocked, and Prettier and ESLint run after every edit. Commits are scanned by gitleaks.

The secret-blocking hook (.claude/hooks/block-secrets.mjs) is a best-effort extra layer: a regex over a shell command can never catch every form. The real boundary is that there are no secrets in the repo to leak. Secret files are gitignored, gitleaks scans every commit, and runtime secrets live in AWS SSM Parameter Store. Every Bash aws command needs explicit approval.

License

MIT

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    B
    maintenance
    Provides semantic search and listing over a unified personal knowledge graph aggregating LinkedIn, GitHub, course completions, and more, enabling MCP clients to answer questions about skills and experience with evidence-backed results.
    1
    MIT
  • F
    license
    Not graded
    quality
    C
    maintenance
    Enables AI assistants to query a person's CV and portfolio content via MCP tools and resources, returning grounded answers from local markdown data instead of relying on resume parsing.
    -
  • F
    license
    Not graded
    quality
    B
    maintenance
    Enables AI agents to retrieve a personal biography, skills, projects, and contact links through a read-only MCP server.
    -
  • F
    license
    A
    quality
    C
    maintenance
    A local MCP server that exposes personal data from JSON files via tools, resources, and prompts, enabling profile reading, project/experience lookup, data search, and bio generation.
    8
    -