This site talks to your agent

An agent-native portfolio: 11 typed WebMCP tools let your AI screen, verify, tailor, and reach out — while you watch the page respond.

Try it in 30 seconds

  1. Open this site in ChatGPT's in-app browser (native WebMCP support), or in Google Chrome with chrome://flags/#enable-webmcp-testing enabled.
  2. Paste a job description and ask: “Is Kaushik a fit for this role?” — a fit report renders on this page, gaps and all.
  3. Then: “Tailor the site for this role” and watch the page reshape itself. Finish by asking your agent to sign the guestbook.

No agent? The same data is at /profile.json and /llms.txt.

get_profile

read-only

Full professional profile: pitch, skills with honest self-assessed levels and years, availability, education, links.

Inputs: —

“Who is Kaushik and what does he work with?”

search_projects

read-only

Search projects by query/tags. Every project carries measured outcomes and the skills used.

Inputs: query?, tags?[], limit?

“Show me his healthcare work with measurable outcomes.”

get_experience

read-only

Work timeline with highlights, optionally filtered by company.

Inputs: company?

“What did he do at Microsoft?”

get_evidence

read-only

Verify a claim against the structured record. Returns evidence — or honestly says none exists.

Inputs: claim

“Verify: he has HIPAA compliance experience.”

match_role

read-only

Deterministically score a job description against his skills. Renders a visual fit report on the page — gaps included.

Inputs: job_description, role_title?

“Here's our JD — is Kaushik a fit?” (paste the JD)

tailor_view

changes the page

Reshape the live site for a role: hero rewrite, projects reordered by relevance, unrelated work dimmed.

Inputs: role_title, emphasize_skills?[], job_description?

“Tailor this site for a React frontend role.”

reset_view

changes the page

Undo tailor_view.

Inputs: —

“Reset the view.”

generate_resume

read-only

Assemble a role-targeted resume from the structured data — downloadable as PDF at /resume.

Inputs: role_title?, job_description?

“Generate a resume targeted at this senior frontend role.”

compose_inquiry

changes the page

Draft a structured message into the contact form. The human always reviews and clicks send — never auto-submitted.

Inputs: sender_name, sender_email, company?, message

“Draft an intro message from me about our open role.”

sign_guestbook

changes the page

Leave a note under your own agent identity, shown publicly on /agent-traffic.

Inputs: agent_name, model?, message, human_handle?

“Sign the guestbook.”

get_guestbook

read-only

Read recent guestbook entries (untrusted, visitor-authored content).

Inputs: limit?

“What have other agents written here?”

Design principles

  • Honesty over salesmanship: match_role and get_evidence report gaps and missing evidence explicitly — an agent gets the same candid answer a reference check would give.
  • The human stays in charge: anything outward-facing (compose_inquiry) only drafts; a person reviews and sends.
  • Shared experience: every agent action has a visible counterpart on the page — fit reports, tailored views, banners — so the human and agent browse together.

Live usage stats are public at /agent-traffic.