Capture from Every AI Gateway
Drop-in capture middleware for the platforms you already use. One configuration change. No agent instrumentation, no prompt modification.
Capture Layer Adapters
First-class capture adapters for every major AI gateway platform.
Express Middleware
@skills-trace/express
bash
npm install @skills-trace/expresstypescript
import { createMiddleware } from '@skills-trace/express';
app.use('/v1/chat/completions', createMiddleware({
registry: './fingerprints.json',
emitter: 'jsonl:./telemetry.jsonl'
}));Vercel AI SDK
@skills-trace/vercel
bash
npm install @skills-trace/verceltypescript
import { skillsTrace } from '@skills-trace/vercel';
const middleware = skillsTrace({
registry: './fingerprints.json'
});Cloudflare Workers
@skills-trace/cloudflare
bash
npm install @skills-trace/cloudflaretypescript
import { createHandler } from '@skills-trace/cloudflare';
export default { fetch: createHandler({ /* config */ }) };LiteLLM Python
skills-trace (PyPI)
bash
pip install skills-tracepython
from skills_trace import SkillsTraceCallback
litellm.callbacks = [SkillsTraceCallback()]Feature Comparison
All adapters feed the same Capture Layer and downstream curation pipeline. Here is what each one supports.
| Feature | Express | Vercel | Cloudflare | LiteLLM |
|---|---|---|---|---|
| Capture telemetry | ✓ | ✓ | ✓ | ✓ |
| Fire-and-forget | ✓ | ✓ | ✓ | ✓ |
| Streaming support | ✓ | ✓ | ✓ | — |
| Multi-tenant routing | ✓ | ✓ | ✓ | ✓ |
| Redaction-by-default | ✓ | ✓ | ✓ | ✓ |
| Custom emitters | ✓ | ✓ | ✓ | ✓ |
| TypeScript | ✓ | ✓ | ✓ | — |
| Python | — | — | — | ✓ |