Tiers
"Infrastructure is free. Intelligence is the product."
CQ has three tiers. Free gives you everything to get started — local memory, full MCP toolset, and one GPU. Upgrade to Pro when you want your AI to remember across projects and platforms. Add Team for shared GPU infrastructure and collective knowledge.
Comparison
| Feature | Free | Pro | Team |
|---|---|---|---|
| Price | $0 | $36/yr (Early Bird) | $36/seat/yr |
| Project-local memory | Yes | Yes | Yes |
| Cross-project memory | No | Yes | Yes |
| Cross-platform sync (Claude↔ChatGPT↔Cursor) | No | Yes | Yes |
| All MCP tools | Yes | Yes | Yes |
| E2E encrypted relay (NAT traversal) | Yes | Yes | Yes |
| GPUs | 1 (local only) | Unlimited | Unlimited (pooled) |
| AI autonomous experiment loop | No | Yes | Yes |
| Persona learning (AI codes like you) | No | Yes | Yes |
| Team knowledge auto-sharing | No | No | Yes |
| Team GPU pooling | No | No | Yes |
| Privacy isolation between orgs | No | No | Yes |
| Billing | — | Annual only | Annual only |
Early Bird: Pro is $36/yr for the first 1,000 users, locked forever at that price. After the first 1,000, pricing will increase.
Free — $0
Everything runs locally. Full MCP toolset. One GPU worker (your local machine or one attached server).
# No login required. Configure your own LLM keys.
cq secret set openai.api_key # Stored encrypted in ~/.cq/secrets.db# .c4/config.yaml
llm_gateway:
enabled: true
default: openai
providers:
openai:
enabled: true
default_model: gpt-4o-miniWhat you get:
- Full MCP tool suite (knowledge record/search, task orchestration, relay)
- Project-local knowledge base (SQLite, stays on your machine)
- Knowledge Loop within a single project
- E2E encrypted relay for NAT traversal
- 1 GPU worker
What you don't get:
- Knowledge doesn't follow you to a second project
- Knowledge doesn't sync to ChatGPT, Cursor, or other AI tools
- No persona learning across tools
Good for: offline use, air-gapped environments, trying CQ before committing.
Pro — $36/yr (Early Bird, first 1,000 users)
Your AI memory becomes persistent, cross-project, and cross-platform. Connect unlimited GPUs.
cq auth login # GitHub OAuth, one-time
cq serve # Start relay + knowledge sync + token refreshWhat you get on top of Free:
- Cross-project memory — knowledge from project A surfaces in project B automatically
- Cross-platform sync — discover something in Claude Code, use it in ChatGPT the next day
- Unlimited GPUs — connect any machine via
cq serve; it joins your GPU pool - AI autonomous experiment loop — CQ plans, submits, evaluates, and iterates without you watching
- Persona learning — AI learns your coding style, preferred patterns, and judgment over time
Trigger to upgrade from Free:
- You run
cq import chatgptand want phase 2–3 (cross-platform sync) - You start a second project and want knowledge to carry over
# After login, connect a GPU machine (any machine, any network):
ssh gpu-server "cq serve" # That machine joins your GPU pool
cq gpu status # See it appear
# GPU: gpu-server (RTX 4090, 24GB) — online, idleAnnual billing only. No monthly option.
Team — $36/seat/yr
Everything in Pro, plus collective intelligence and shared infrastructure for your team.
# Team admin setup:
cq auth login
cq team create my-org
cq team invite alice@company.com bob@company.com
# Each member runs:
cq auth login
cq serveWhat you get on top of Pro:
- Team knowledge auto-sharing — when one teammate's AI discovers a pattern, it surfaces in everyone's context automatically
- Team GPU pooling — all connected machines form a shared GPU pool; submit jobs from anywhere, run on any available GPU
- Privacy isolation — your org's knowledge is isolated from other orgs; no cross-org leakage
# Submit a training job from your laptop, run on team GPU:
cq job submit --script train.py --gpu any
# Submitted job-a4f2 → assigned to alice-rtx4090 (available)
# Streaming logs...When Alice's AI learns "always check gradient norms before epoch 5", that pattern propagates to your AI's context — skipping the trial-and-error that Alice already paid for.
Annual billing only. Per seat.
What Tier Am I On?
cq auth statusTier: Pro
User: changmin@pilab.co.kr
Relay: connected (relay.pilab.kr)
Knowledge: cloud (pgvector, 847 records)
GPUs: 2 connected (local + gpu-server)Tier is determined by your login state and whether cq serve is running:
- Free: Not logged in, or logged in but running fully offline
- Pro: Logged in +
cq serve+ no Team configuration - Team: Logged in +
cq serve+ team workspace configured
Data stored locally (Free) is preserved when you upgrade. Cloud sync picks up from where local left off.
Knowledge Loop (all tiers)
The Knowledge Loop runs on every tier, but scope differs:
| Tier | Knowledge scope |
|---|---|
| Free | Single project only |
| Pro | All your projects + all your AI tools |
| Team | All your projects + all AI tools + teammates' discoveries |
See Knowledge Loop for how the learning cycle works.
Research Loop (Pro and Team)
Run autonomous ML experiments on your connected GPUs without watching:
cq research run --goal "maximize accuracy on CIFAR-10" --budget 20Loop started: research-7f3a
Iteration 1: submitting experiment...
Training on gpu-server (RTX 4090)...
Result: acc=0.847, loss=0.421 @epoch=30
Planning iteration 2 based on results...
Iteration 2: adjusting learning rate schedule...
Result: acc=0.863, loss=0.391 @epoch=30
...
Best checkpoint saved to Drive: ckpt-research-7f3a-iter8.ptThe loop runs until it hits the budget or a stopping condition. Results stream back through the relay.