AI the right way.
Every AI has blind spots. 1K4 catches them.
Five named advisors review your code, plans, and decisions in parallel and surface where they disagree. That's Quorum, the 1K4 peer-review layer for Claude Code, Codex, Cursor, terminal, or 1K4 Agent Lab.
Setup is free. Cloud reviews run on prepaid credits with a per-run cap.
Two review tools, one layer: Quorum for open-ended questions. 1k4-review for plan-mode checks before file changes.
Paste this to your coding agent.
Your agent installs 1K4 and signs you in via the browser. Safe to re-run.
Install the 1K4 CLI (verify: https://www.npmjs.com/package/@1key4ai/cc-bridge). Run: npx -y -p @1key4ai/cc-bridge@1.9.12 cc-bridge login. If npx is missing, install Node.js LTS and retry. It adds a 1K4 entry to my coding-agent configs, keeping existing settings; it won't ask me to paste an API key or password. If it shows a URL and code, relay both, wait for approval, then rerun the same command in this chat. Eligible new accounts get $5 free credit after their first successful plan review. Done when it prints "Signed in."No token to copy · Claude Code · Codex · Cursor
Terminal review running
See a review
This is what a Quorum review looks like.
Three example advisors from Quorum's five. Switch tabs to compare how each output shape differs on a real prompt.
You: Use the Code Auditor on backend/auth/middleware.py. Worried about session fixation.
Critical
Session ID not regenerated post-login. An attacker who plants a known session ID before login retains access after authentication. Evidence:
middleware.py:47 (set_user_id updates payload but never calls session.regenerate_id()).Important
No CSRF token check on state-changing routes. Logout and password-change handlers accept POST without verifying a CSRF token. Evidence:
middleware.py:71-86.Advisory
Open redirect via
next query param. Post-login redirect honors any URL passed in ?next=. Evidence: middleware.py:102; add ALLOWED_HOSTS check.Next step: Run this on your own repo.Get started →
How peer review works
Independent answers, then disagreement, then a decision you can act on.
1
Ask
Send a plan, a question, or a decision. One model would just answer.
2
Separate
Multiple models respond independently. No copying, no echo chamber.
3
Challenge
The models review each other. Disagreement is surfaced, not hidden.
4
Decide
You get consensus, dissent, and a recommended next step. Use it where you work.
Multi-model review is a pressure-test, not a guarantee. Models can share blind spots. We surface disagreement so you can judge.
Use it where the work happens
Three lanes, one workflow.
For Claude Code and Codex users
Review a plan before you implement it.
Hand Claude Code's plan to other models before any files change. Catch missing migrations, missed edge cases, security gaps. Promo lives on the Plan Reviewer page.
See plan reviewer →For architects and operators
Ask Quorum a hard question.
Five named specialists: Code Auditor, Research, Analysis, Counterpoint, Budget. Ask one. Or run a brainstorm and watch them argue.
Meet Quorum →When work grows
Continue the review in Lab.
Files, sessions, diffs, deploys, multi-turn agent work. Lab is the workspace when one review isn't the whole job.
Open Lab →Inside 1K4 Lab
The workspace where review turns into work.
Peer review gets you to a decision. Lab is where that decision turns into files, revisions, deploys, and follow-up work in the same project.
START WITH DIRECTION
Brainstorm and review in one place.
Turn a rough brief into a structured plan, then pressure-test it before work begins.
Brainstorm + ReviewMOVE INTO ARTIFACTS
Build outputs you can inspect.
Create mockups, revise documents, and generate code in a workspace that keeps edits visible.
BuildSHIP FROM CONTEXT
Deploy from the same session.
Provision, build, and publish without losing the thread that produced the decision.
DeployKEEP WORK MOVING
Automate the handoffs.
Track queued tasks, watch progress, and step in only when the agent needs a decision.
AutomateCONNECT THE REST OF 1K4
Code, models, and apps stay attached.
Move from Lab into code review, model routing, and connected apps without changing projects.
Code workspace + Routing + API→
Tour Lab
Public tour. The member workspace opens after setup.
Pricing
Free setup. Capped cloud reviews.
Connect local models for $0. Cloud reviewers run on prepaid credits, with a per-run cap before anything starts.
FREE
Your own models
$0
Connect your local Ollama via the 1K4 Bridge. Full agent capability on your hardware. Performance depends on your model.
CLOUD
Cloud models
$1 = 1 credit
Buy credits. Use any model from any provider. Per token, per image, per second. No overage. No expiry. Set a per-run cap so a single Quorum or review cannot overspend. Typical Quorum runs cost $0.10 to $2.00.