Supported Agent

Claude Code

Anthropic's terminal-based AI coding agent. Send a session log and Videlic runs all four analyzers — honesty check, ripple risk, QA lens, and full timeline.

What Videlic catches

Every way Claude Code can mislead you

Hallucinated summaries

Claude says it "added tests and refactored auth" — but the test file was deleted and auth.ts has a missing guard. We catch the gap.

Silent high-risk edits

A one-line change in a shared utility with 14 importers. Claude won't tell you about the blast radius — we will.

Fake test coverage

expect(true).toBe(true) passes every CI run but tests nothing. Videlic flags assertions that can never fail.

How to connect

One command to analyze

# Send your Claude Code session to Videlic

$ videlic analyze ~/.claude/projects/*/session.jsonl

→ Analyzing... 4 checks running

✓ Report ready → videlic.dev/s/abc123

1

Run your session

Use Claude Code as you normally would. Sessions are logged to ~/.claude/ automatically.

2

Send to Videlic

One CLI command. Agent format auto-detected, zero configuration needed.

3

Get your report

A shareable URL with trust score, discrepancies, and actionable insights. Ready in seconds.

Example report

claude-code·feat/auth-refactor
62/100
🔍

Honesty Check

3 discrepancies

Claimed "updated all tests"

auth.test.ts deleted, not updated

🌊

Ripple Risk

High — 14 deps

src/lib/auth.ts modified

middleware chain → 6 route handlers

🧪

QA Lens

2 fake assertions

expect(result).toBeDefined()

passes regardless of value

Timeline

47 operations · 12 min

4 rewrites, 2 test failures

retried auth.ts 3 times

src/lib/auth.tssrc/middleware/session.tsauth.test.ts+5 files