Documentation
Everything you need to integrate and operate Skalpel AI.
Getting Started
Quick Start
Go from zero to first proxied request in under 10 minutes.
- 1. Create a workspace at skalpel.ai
- 2. Generate your workspace key
- 3. Run: npx skalpel init
- 4. Send your first request
Base URL Swap
The fastest integration path. Keep your existing SDK.
- Replace your provider base URL with https://gateway.skalpel.ai/v1
- Set your SKALPEL_KEY environment variable
- Send traffic as normal
SDK Wrapper
Lightweight wrapper for OpenAI and Anthropic clients.
- npm install @skalpel/sdk
- Wrap your existing client instance
- Keep your existing code intact
CLI Reference
skalpel init
Scan repo, detect SDKs, patch safely.
skalpel connect
Connect workspace and verify credentials.
skalpel doctor
Run diagnostics on your integration.
skalpel logs tail
Stream live request logs from your workspace.
API Reference
POST /v1/chat/completions
OpenAI-compatible chat endpoint.
POST /v1/messages
Anthropic-compatible messages endpoint.
GET /api/workspaces
List workspaces.
GET /api/requests
Query request history and traces.
GET /api/billing
Current usage and savings breakdown.
Concepts
Request Lifecycle
Ingress, normalization, auth, policy resolution, cache, routing, forwarding, tracing.
Savings Accounting
Verified, estimated, and provider-side savings buckets.
Quality Floors
Shadow evaluations and kill switches for every optimization.
Provider Adapters
Normalized interface across OpenAI, Anthropic, Google, and more.