pricing
Priced per credit. Net-positive by design.
One credit costs $0.05 and covers roughly 100 MB of tool output — about $0.50 per GB, a small fraction of the model tokens that raw output would burn to read. Start with 10 free credits (~1 GB), buy more only when you need them, and results pass through unchanged.
Start free. No card until you buy.
- 10 free credits (~1 GB) to start
- 1 credit ≈ 100 MB ≈ $0.05
- Packs: ~1 GB / $0.50 · ~10 GB / $5 · ~50 GB / $25
- Custom credit packs anytime
- Runs out → optimization pauses, no surprise bills
- Claude Code, Codex, CLI, API and MCP
For teams running agents in anger.
- 12,500 credits / mo (~1.25 TB) — ~20% more than pay-as-you-go
- Overage at the standard $0.50 / GB
- Inference pipeline for CLI, API and MCP
- Per-member usage attribution
- Priority support · 24h SLA
- Dedicated Slack channel
SSO, regulated data, custom SLAs.
- Contractual volume
- SSO / SAML
- VPC peering
- Client version pinning
- Custom SLAs
- Dedicated Slack channel
Dollar savings shown in the product are estimates from public model prices.
credits
One credit, five cents. Each tool spends it at its own rate.
A credit costs $0.05. You spend credits only on the tool output Codag reduces, and each tool burns them at a published rate: log compaction is cheap (~100 MB per credit, about $0.50/GB), while inference-heavy agent-context reduction runs a frontier model and burns credits far faster (~100 KB per credit). MB/KB are estimates; the credit is the billed unit. Every credit is designed to return more than it costs.
| Tool | Credit rate | ≈ per GB |
|---|---|---|
| Log compaction | 1 credit / ~100 MB | $0.50 / GB |
| Agent-context reduction (inference) | 1 credit / ~100 KB | ~$500 / GB |
1 credit = $0.05. Buy in packs — ~1 GB of logs for $0.50, ~10 GB for $5, ~50 GB for $25. Different tools spend the same credits at different rates: logs are cheap (deterministic), inference reduction runs a frontier model and burns credits ~1000× faster. MB/KB are estimates; the credit is the billed unit. Your first 10 credits (~1 GB of logs) are free — a one-time starter grant, no card required.
faq
Credits, answered
What is a credit?
A credit costs $0.05 and is the single unit across the CLI, API, and MCP. Each tool spends credits at its own published rate: log compaction burns ~1 credit per 100 MB (about $0.50/GB), while inference-heavy agent-context reduction runs a frontier model and burns ~1 credit per 100 KB (~1000× faster). MB/KB are estimates; the credit is the billed unit, and you spend them only on the tool results Codag actually reduces — never on seats.
What happens when I run out of credits?
Optimization pauses and your agent keeps getting the original tool output byte for byte — nothing breaks. There are no surprise overage bills; you top up whenever you want, by the credit.
Do the free credits expire?
No. The 10 free credits (~1 GB) are a one-time starter grant so you can try Codag with no card. Once they're used you continue at $0.05 per credit, or buy a pack.
How is Team different from pay-as-you-go?
Team is $499 per month for a monthly bundle of 12,500 credits (~1.25 TB) — about 20% more credits than buying pay-as-you-go, with any overage billed at the standard $0.50 per GB. It adds the inference pipeline across CLI, API, and MCP, per-member usage attribution, priority support, and a dedicated Slack channel. Pay-as-you-go bills purely by the credit — $0.05 each, about $0.50 per GB — and is the fastest way to start; Team is for teams running agents every day.
Why is Codag net-positive? What's the ROI?
A credit costs $0.05 and compresses roughly 100 MB of tool output — about $0.50 per GB — but that raw output of logs, test output, or search results usually costs your agent far more than that in model tokens to read. You pay cents to save dollars — the ROI is the whole point.