Secrypt
Private AI API Guide: OpenAI-Compatible Cipher
A private AI API lets agents and IDEs call chat completions without handing your prompts to a training-heavy consumer stack. Cipher is Secrypt’s OpenAI-compatible surface.
Key takeaway: A private AI API lets agents and IDEs call chat completions without handing your prompts to a training-heavy consumer stack. Cipher is Secrypt’s OpenAI-compatible surface.
Why developer API demand is evergreen
Agents, IDE copilots, and automation will keep needing chat completions. The durable question is which backend sees those prompts — and whether that backend trains on them or meters them honestly.
OpenAI-compatible shapes persist because SDKs and tools standardized on them. A private endpoint that speaks the same dialect wins migrations.
Cipher basics
Base URL: https://secrypt.space/v1. Model id: cipher (alias secrypt-cloud). Auth: Bearer sk_sec_… from Account → API keys.
Call POST /v1/chat/completions like any OpenAI-shaped client. Docs live at /docs; product overview at /api.
| Plan | API included / day | Then |
|---|---|---|
| Free | 50 | $0.01 / request credits |
| Pro | 2,000 | $0.01 / request credits |
| Unlimited | 10,000 | $0.01 / request credits |
Chat meters ≠ API meters
Secrypt separates chat UI quotas from API key quotas. Cursor can burn API allowance while your browser chat still has messages left — and the reverse.
Chat overage is $0.05/message via credits. API overage is $0.01/request. Buy credits from Account when agents are busy.
IDE and agent wiring
Point Cursor, Continue.dev, or VS Code OpenAI-compatible backends at Secrypt’s base URL with your key. Prefer short system prompts and redact secrets in tool outputs.
Revoke keys when a laptop is lost. Rotate keys per environment (dev vs prod).
- Never commit sk_sec keys to git
- Use separate keys per machine or CI job
- Watch Account usage for chat vs API
- Read /blog/use-secrypt-in-cursor for a cookbook
Local LLM vs private hosted API
Local wins when air-gapping is mandatory. Private hosted API wins when you want less ops and still care about no-training posture. Many teams use both: local for sealed code, Secrypt Cipher for everyday agent work.
Deep-dive guides in this cluster
Related posts cover OpenAI-compatible Cipher details, Cursor setup, Continue.dev, VS Code backends, and API pricing explained.
Frequently asked questions
Is Cipher the same as ChatGPT’s API?
It is OpenAI-compatible in request shape. It is Secrypt’s private Cipher model path with Secrypt privacy and pricing — not OpenAI.
Do API calls train Secrypt models on my prompts?
Secrypt’s product posture excludes conversation content from model training and ad profiling. Read /privacy for the full policy.
What model id should I send?
Prefer cipher. secrypt-cloud is an alias.
How do I start?
Create an account, mint a sk_sec key, set base URL to https://secrypt.space/v1, and run a tiny chat completions smoke test.
Is there unlimited free API?
No. Free includes 50 API requests/day, then pay-as-you-go credits. Paid plans raise the daily allowance.
Related deep-dives
Try Secrypt
Secrypt is private, uncensored AI chat with an OpenAI-compatible Cipher API. No training on your messages.