For AI agentsThe resume engine your agent can call.
Paste one URL into Claude or any MCP client and sign in. From then on your agent can lint, match, tailor, and save resumes on your Zelume account. No API keys, no config files, nothing to copy but the address.
https://www.zelume.io/mcp
Claude · zelume connected
you Tailor my resume for this posting: stripe.com/jobs/platform-eng
→ fetch_posting ✓ · match_jd 48%
missing: Kubernetes, Terraform, on-call
Your Platform resume tells the infra story but never says Kubernetes or Terraform, and both are in your last two roles. Want me to work them in and save a tailored copy?
you Do it
→ lint_resume 91/100 · push_my_resume ✓
saved as “Stripe · Platform Eng”
ConnectOne address, whatever you run
Claude.ai & Claude Desktop
Add a connector
Settings → Connectors
Add custom connector
https://www.zelume.io/mcp
Sign in when the browser opens
No client ID, no advanced settings. Paste the URL and the login screen does the rest.
claude mcp add --transport http \
zelume https://www.zelume.io/mcp
/mcp # then authenticate
The browser opens once for login. After that the tools show up in every session.
~/.cursor/mcp.json
{ "mcpServers": { "zelume":
{ "url": "https://www.zelume.io/mcp" } } }
Save the file, then click the sign-in prompt Cursor shows next to the server. Keep Cursor updated; recent builds handle the login flow cleanly.
code --add-mcp '{"name":"zelume",
"type":"http",
"url":"https://www.zelume.io/mcp"}'
For agent mode in VS Code 1.101 or newer. Approve the sign-in prompt and the browser handles the rest.
Any MCP client
Standard everything
transport: streamable HTTP
auth: OAuth 2.1 + PKCE
registration: automatic (DCR)
discovery via .well-known
Clients register themselves through RFC 7591, so there is nothing to pre-configure. A client that can’t do OAuth can send a personal access token as a bearer instead.
Sign-inThe login is the whole setup
Your client reads the server’s discovery metadata, registers itself, and sends you to auth.zelume.io. You sign in the same ways you do on the site: password, Google, or a one-time email code.
Every connection gets its own credential
Each client registers as its own OAuth client and holds its own token. Nothing is shared between your connections, or with anyone else’s.
Scoped away from your web session
Bridge access requires a dedicated OAuth scope that web-app sessions never receive. A leaked browser token can’t drive the bridge, and a bridge token can’t do anything the tools don’t expose.
Leaving is one click
Remove the connector in your client and the access goes with it. Personal access tokens, if you use those instead, revoke from Settings.
The toolsWhat your agent can do
Engine tools
Deterministic. The same input always gets the same answer.
lint_resumeATS lint with a 0–100 score. Every deduction has a named rule behind it.
match_jdKeyword-gap analysis between a resume and a job description, pasted or by URL.
validate_resumeChecks a document against the JSON Resume schema.
fetch_postingPulls a job posting server-side, with SSRF guards on where it will go.
pingHealth check.
Account tools
Available once you sign in. Deliberately narrow.
list_my_resumesLists your resumes. Metadata only, never content.
fetch_my_resumeFetches one resume’s full content by id.
push_my_resumeSaves a new resume to your account. Create only: it cannot edit or delete existing ones.
export_pdfRenders a resume to PDF through your account templates.
list_templatesLists the templates you can render with.
lookup_termLooks up an allowlisted term on Wikipedia or npm, for fact-checking claims.
The account surface is small by design. Listing returns metadata. Pushing creates and never overwrites. There is no delete, no settings access, no billing access. If a tool isn’t on this page, your agent doesn’t have it.
The agent does the legwork. You stay the author.
Scores come from named rules, match percentages from keyword math. When the engine says 48%, a second run says 48%. Your agent gets honest numbers to work with, and what it saves lands next to your other resumes for you to review before anything goes out.
First movesThings to ask once it’s connected
>Tailor my resume for this posting: <paste a URL>. Show me the keyword gaps before you change anything.
>Which of my resumes fits this job description best?
>Lint my resume and walk me through every deduction.
>Save the tailored version to my account as “Acme · Platform Eng” and export a PDF.
One address. That’s the setup.
https://www.zelume.io/mcp
A free account is enough to connect. If you’d rather keep everything on your own machine, the same engine ships as an open-source CLI with a local agent bridge.