v0.5.0 — early access

Human signal, on demand.
Native to your agent workflow.

One command. Real feedback. Signal in minutes.
No forms. No dashboards. Just your terminal.

$curl -fsSL https://candor.sh | bash
Works seamlessly with
Claude Code
Gemini CLI
Codex
terminal
$ claude "rank these outputs by quality"
▶ candor study running...
Ranking 3 outputs across 24 responses...
Output C ranked #1 by 92% of participants
Output A ranked #2 — strong on clarity
Output B flagged for tone issues
your terminal
feedback
★★★★☆4.2 avg rating
#1Output C (92%)
#2Output A (87%)
human feedback
terminal
// how it works
1
install
One curl command. Registers itself into Claude Code automatically.
2
create a study
Tell Claude what you want to learn. It sets up the study — you just approve.
3
share or recruit
Share the link yourself, or let Candor recruit participants from a research panel automatically.
4
responses come in
Participants complete tasks or join moderated sessions. Results stream in real time.
5
get results
Ranked outputs, labeled data, or prioritized findings. Feed them into your training pipeline or back to your agent.

What participants see

Every interface is optimized for focus and high-fidelity feedback.

Pairwise Comparison
A is better
Tie
B is better
Rating Scale
1
2
3
4
5
Categorization
Positive
Neutral
Negative
Free Text
Describe what you see in this image...

AI-moderated voice interviews

Participants browse your product while an AI voice moderator asks questions, probes on friction points, and adapts follow-ups in real time.

Hear a demo →
Voice Interview
CANDOR
Welcome to your study session

This session uses an AI voice guide. Please find a quiet place and ensure your microphone is working before we begin.

Allow microphone & Begin
Powered by Candor

Ready to get human signal?

Install Candor and start collecting feedback in minutes.

$curl -fsSL https://candor.sh | bash
>