Messages and Count Tokens
Claude Messages, message token counting, and a Claude Chat Completions-compatible endpoint.
Coding Plan · Multi-model codingOne plan.
One plan.
Multiple coding models.
One token for Claude Code, Codex, Kimi Code, and more. Choose models by task and track usage in one place.
Multiple models for coding tasks
Access 10 public APIs across Messages, Responses, and Chat Completions. Model availability depends on the live directory, credential permissions, and client configuration.
Responses and Chat Completions
The Responses protocol used by Codex and OpenAI Chat Completions.
Multi-model Chat Completions
Public Chat Completions endpoints for Gemini, GLM, Grok, Kimi, and DeepSeek.
More Credits for a limited time, with no price increase
The limited-time offer is scheduled to end September 30, 2026; the checkout page is authoritative for eligibility and Credits received. Every tier has the same Coding model access and differs only in included Credits.
Coding Plus
500 Credits400 base + 100 bonus · ¥99For short projects, test coverage, and occasional use.
Coding Standard
1,600 Credits1,200 base + 400 bonus · ¥299For everyday development, multiple repositories, and code review.
Coding Pro
4,000 Credits3,000 base + 1,000 bonus · ¥699For sustained agents, complex refactors, and frequent use.
Coding Max
10,000 Credits8,000 base + 2,000 bonus · ¥1,399For team workloads, multi-agent runs, and automation.
Transparent usage instead of vague request counts
Actual consumption depends on the model, context, cache behavior, output, tools, and parallel work.
Four separately metered usage types
- Uncached input
- Model output
- Cache writes
- Cache reads
Purchase and balance boundaries
- Renew before expiry to carry forward remaining Credits and extend the term by 30 days
- Unused Credits expire at the end of the term; renewing afterward grants only the new package amount and starts a new 30-day term
- No automatic renewal
- No fallback to the general balance
- Normal API concurrency and rate limits still apply
Choose the protocol that matches your client
The setup center provides working configurations for Claude Code, Codex, Kimi Code, OpenClaw, Hermes, Kilo Code, Aider, Qwen Code, and DeepSeek Harness. Each guide pins the matching endpoint, authentication method, protocol, and launch command so you can choose directly by client.
Frequently asked questions
Do tiers have different model permissions?
No. All tiers have the same Coding model access. Exact models depend on the live directory and the compatibility range of your client profile.
How are Credits deducted?
Uncached input, output, cache writes, and cache reads are metered separately. Long context, tools, and parallel tasks can increase consumption.
Will Coding use my general balance automatically?
No. Coding Applications disable general-balance fallback. You decide whether to purchase again.
How do I see current models?
Request the live model directory with a valid Coding credential and use an exact model ID marked compatible by your current profile.
How do I troubleshoot a configured client?
First confirm that the token belongs to a Coding Application, then compare the endpoint, protocol, model ID, and launch method with the client guide.
Choose a plan, then choose the model for each task
One token, one bill, and per-call usage details. Start with the guide for your client.
Introduction
APIs
Pricing