interviewsignal 0.1.5

AI-native interview platform — capture thought process, not puzzle performance

interviewsignal

An AI-native interview platform. Type /interview in Claude Code, Codex, Cursor, or any AI coding assistant — it captures your entire thought process as you solve a problem and sends a structured, tamper-evident audit to the hiring manager.

No contrived puzzles. No whiteboard anxiety. No bias. Just signal.

---

The problem

You're hiring a software engineer. You've spent 6 hours watching three candidates struggle with binary tree problems none of them will ever encounter on the job. One froze. One solved it but couldn't explain their thinking. One was brilliant but had a bad day.

You still don't know who can actually build software.

Meanwhile, every one of those candidates uses AI coding assistants every day. You tested them without their tools — like testing a surgeon without instruments.

There's a better signal: how they think.

---

What you get

For hiring managers:

• A timestamped audit of everything the candidate did — every prompt, every tool call, every file written
• AI grading against your own rubric (not a canned scoring system)
• Anonymous-first dashboard: you see scores before names, every time
• Grade-before-Reveal enforcement with a tamper-evident audit trail — provable merit-first hiring
• Comment threads, hire/reject decisions, all hash-chained and email-anchored

For candidates:

• Work the way you actually work — with AI assistance, in your own environment
• Get evaluated on your thinking, not your ability to memorise algorithms
• No file transfers, no email attachments — just a short code

For teams:

• Close the interview loop in half the time
• A written record of the hiring decision from problem to offer
• Works inside your existing toolchain — no new platform to log into

---

Install

pip install interviewsignal && interview install

Requires Python 3.10+ and one of: Claude Code, Codex, Cursor, Gemini CLI, Aider.

Then configure grading and (optionally) the relay:

interview configure-api-key    # Anthropic API key — for grading
interview configure-relay      # relay URL — auto-registers your HM account

Enterprise / no personal API key? See Enterprise configuration below.

---

Quickstart

Hiring manager

/interview hm

You'll be asked for:

• Problem statement
• Grading rubric (plain language — "weight decomposition 40%, code quality 30%, tests 30%")
• Your email + CC list (HR, co-interviewer, etc.)
• Audit recipient (HR or a neutral third party — receives a silent email on each key action)
• Time limit (optional)
• Anonymize candidates? (yes / no)

You get back a code like INT-4829-XK. Share it with your candidate — that's all they need.

Candidate

pip install interviewsignal && interview install

/interview INT-4829-XK

The problem appears. Relay is auto-configured — no API keys or file transfers required. Work normally — ask the AI questions, write code, run tests. The session records everything automatically.

When done:

/submit

The session is sealed and sent to the relay. The hiring manager's dashboard updates automatically.

Hiring manager — review

interview dashboard
# → http://localhost:7832

Candidates appear as "Candidate A", "Candidate B" — scores first, names second. Click into any candidate to see the full transcript (prompts + AI reasoning + tool calls), dimension scores, and diff. Add comments. Record your decision. Click Reveal when you're ready to unmask. Use Verify Chain to confirm the session log is tamper-evident.

---

How it works

interviewsignal installs as a skill into your AI coding assistant. It captures the full conversation — the candidate's prompts, the AI's reasoning before each action, every tool call (reads, writes, bash commands) — and builds an append-only, hash-chained session log. On /submit, the log is sealed and pushed to the relay. The HM grades from their dashboard using their own AI key.

Candidate side                          HM side
─────────────────────────               ───────────────────────
interview configure-relay               interview configure-relay
  ↓ auto-registered via relay             ↓ gets unique hm_key

/interview hm                           ← share code INT-4829-XK
  ↓ creates interview
  ↓ pushes package to relay

/interview INT-4829-XK                  interview dashboard
  ↓ fetches problem from relay            ↓ localhost:7832
  ↓ relay auto-configured locally         ↓ candidates arrive
Session starts                            ↓ anonymous by default
  ↓ hooks capture every tool call         ↓ Grade All / Grade Selected
  ↓ append-only events.jsonl              ↓ score before name
  ↓ hash chain (tamper-evident)           ↓ Reveal unlocks after grading
/submit                                   ↓ comment thread
  ↓ session sealed                        ↓ hire / next round / reject
  ↓ pushed to relay                       ↓ full audit trail

Three passes on submit:

1. session seal — finalises hash chain, captures git diff (start → end)
2. Push to relay — sealed session (events + manifest + report) stored server-side
3. HM grades from dashboard — sends timeline + rubric + diff to their AI key, returns structured JSON scores

The integrity model:

Every HM action — grading, revealing identity, adding a comment, recording a decision — is logged with a SHA-256 hash chain. Key events are silently emailed to a designated audit recipient (typically HR). The mail server's timestamp is outside the HM's control. Reveal is physically disabled until a grade is saved, ensuring blind evaluation is provable.

[2026-04-13T10:47:22Z] grade_recorded       INT-4829-XK  hash=d4abe5e6
[2026-04-13T10:52:09Z] identity_revealed    INT-4829-XK  hash=2370be19

"Identity revealed 4.8 minutes after grade was recorded." That one line proves merit came first.

---

Relay

The relay stores interview packages and candidate sessions so hiring managers and candidates only need to share a short code — no file transfers, no email attachments.

Run interview configure-relay to choose:

How do you want to deliver interview sessions?
──────────────────────────────────────────────
  1. Your own relay  Railway / Render / self-hosted — private, ~$5/mo
  2. Email only      SMTP — no server, reports arrive by email

Option 1 — Your own relay (~$5/mo, fully private)

Deploy in one click:

After deploying:

1. Set environment variable RELAY_API_KEY (any random string) in Railway → Variables
2. Add a /data volume when prompted — this is where sessions are stored
3. Copy your Railway URL (e.g. https://myrelay.up.railway.app)
4. Run interview configure-relay → option 1 → paste URL

Your data stays in your own Railway account. Cost is ~$5/month (Railway Hobby plan).

Or run it anywhere with Docker:

docker run -e RELAY_API_KEY=secret -v /data:/data -p 8080:8080 \
  ghcr.io/nikhilskashyap/interviewsignal:latest

See docs/self-hosting.md for data layout, backup, and key rotation.

Option 2 — Email only (free, no server)

interview configure-relay   # choose 2
interview configure-email   # set up SMTP credentials

Candidates run /submit → report is emailed directly to the HM. The HM saves the JSON attachment to ~/.interview/received/ and it appears in the dashboard. No server needed, no ongoing cost. Trade-off: manual file handling and the dashboard can't re-grade from raw events.

---

Enterprise configuration

Companies that don't issue personal API keys — or that route all AI traffic through an internal gateway — can configure a custom LLM endpoint:

interview configure-llm

This covers three patterns:

Pattern	What to set
Anthropic direct	API key only (default)
Internal proxy (Floodgate, corporate gateway)	Base URL + optional key; proxy handles auth
OpenAI-compatible endpoint	Base URL + key + format=openai

What gets configured (~/.interview/config.json):

{
  "anthropic_base_url":      "https://ai-gateway.corp.internal/anthropic",
  "anthropic_api_key":       "",
  "api_format":              "anthropic",
  "grading_model":           "claude-3-5-haiku",
  "anthropic_extra_headers": {"X-Team-ID": "ml-hiring"}
}

The API key is optional — leave it blank if your proxy handles auth at the network or SSO level. Extra headers let you pass team/project routing headers required by some gateways.

Environment variable overrides (useful for CI or shared machines):

ANTHROPIC_API_KEY=...           # API key
ANTHROPIC_BASE_URL=...          # base URL override
INTERVIEW_GRADING_MODEL=...     # model name override

---

True meritocracy

Most companies say they hire on merit. Almost none of them do — because the process doesn't allow it.

When you know who someone is before you evaluate them, bias isn't a failure of character. It's a failure of process. The name on the resume, the university on the screen share, the face on the Zoom — they all get into your head before the first line of code is written.

interviewsignal makes meritocracy structurally possible:

Same tools, same environment. Every candidate works with the same AI assistant they use on the job. The person who drilled Leetcode for six months has no advantage over the person who just builds things. The only variable is how well they think.

You can't prevent a candidate from having a second screen. Neither can a Leetcode proctoring tool. The difference is that with interviewsignal, gaming it well requires understanding the problem — and that's the signal.

Score before name, always. Grades are locked in before identity is revealed — not as a policy, but as a technical constraint. You cannot click Reveal until a score is saved. The order of events is cryptographically provable.

A tamper-evident record. Every action — grading, revealing identity, adding a comment, recording a decision — is hash-chained and email-anchored to a timestamp outside your control. The audit trail doesn't just log what happened. It proves it.

[2026-04-13T10:47:22Z] grade_recorded     INT-4829-XK  hash=d4abe5e6
[2026-04-13T10:52:09Z] identity_revealed  INT-4829-XK  hash=2370be19

"Identity revealed 4.8 minutes after grade was recorded."

That one line is the whole argument. You hired the person with the best score. You can prove it. Not because you were careful, but because the system made any other sequence impossible.

---

Platform support

Platform	Install	Hook mechanism
Claude Code (Linux/Mac/Windows)	interview install	PreToolUse + PostToolUse hooks
Codex	interview install --platform codex	PreToolUse hook + AGENTS.md
Cursor	interview install --platform cursor	.cursor/rules/interview.mdc
Gemini CLI	interview install --platform gemini	BeforeTool hook + GEMINI.md
Aider	interview install --platform aider	AGENTS.md

---

What gets captured

Event	Captured
Candidate prompts	Exact message to the AI assistant
AI reasoning	Plan before each action ("I'll use a hash map because...")
File reads	Path
File writes	Path + content hash
Bash commands	Command + exit code
File edits	Path + change summary
Git state	Branch + commit at start and end
Git diff	Full diff (start → submit)
Timestamps	Millisecond precision on every event

The session log is append-only and hash-chained. Any tampering breaks the chain. The HM dashboard includes a Verify Chain button that re-derives every SHA-256 hash and flags any mismatch, along with the relay's server-side submission timestamp.

What is not captured: raw file contents (only paths and hashes, for privacy). The grader evaluates the conversation, timeline, and diff — not file contents.

---

Configuration reference

# Grading
interview configure-api-key    # Anthropic API key (direct access)
interview configure-llm        # Enterprise: custom endpoint, proxy, format, extra headers

# Delivery
interview configure-relay      # Relay URL + auto-register HM account
interview configure-email      # SMTP fallback (no relay)

# Runtime
interview dashboard            # Local HM dashboard at localhost:7832
interview status               # Check active session
interview install --help       # Platform install options

All config stored in ~/.interview/config.json (permissions: 600).

---

Privacy

Candidate sessions are stored on the relay (or locally, in email mode). The relay stores: events.jsonl, manifest.json, report.html, and report.json. Raw file contents are never stored — only paths, hashes, and command summaries.

Grading sends the session timeline and git diff to the configured AI endpoint (Anthropic API by default, or your enterprise proxy). No raw file contents. The grading call uses your own API key — interviewsignal never sees it.

Self-hosted relay: Run the relay inside your own infrastructure and nothing leaves your network. See docs/self-hosting.md.

No telemetry. No analytics. No tracking.

---

Built with

Python stdlib only (no external dependencies for core or relay). Grading via Anthropic Messages API or any compatible endpoint. Dashboard is a self-contained local HTTP server. Reports are single-file HTML. Relay is a single-process stdlib HTTP server backed by flat files.

---

Contributing

Worked examples are the most valuable contribution. Run a real interview session, save the output to worked/{slug}/, write an honest review.md evaluating what the grading got right and wrong, open a PR.

Platform support — each new platform is a ~30 line adapter in cli.py. If you use an AI coding tool not listed above, adding support is straightforward.

See ARCHITECTURE.md for module responsibilities and docs/relay-api.md for the full relay API contract.

---

Thought process, not puzzles. Pure signal.