Integration
One base-URL swap. Everything else is optional.
Ceggesta deploys as a transparent proxy in front of the LLM providers you already use. Your applications keep their SDKs, your keys, your model choices — the only thing that changes is where they point.
Point your SDK’s base URL at Ceggesta. That’s the deployment.
Coverage, by channel
API proxy
The default. One base-URL swap covers your applications’ LLM traffic — attributed, tagged, and streamed back untouched. No code changes, no SDK lock-in.
Browser coverage
A managed extension extends the same intelligence to the consumer surfaces your teams already use — claude.ai, ChatGPT, Gemini, Perplexity — deployed via your device management.
Self-hosted inference
For strict boundaries: Ceggesta’s own analysis models can run inside your environment, so enrichment never leaves your infrastructure.
Design partners with unusual capture requirements can talk to us about additional network-level coverage.
Attribution without SDK work
Five layers, merged per request
Key-level defaults, key-owner identity, per-request headers, your own rules, and source attribution — combined on every event with clear precedence.
Optional headers when you want them
Service accounts can attribute per request with simple X-CG-Tag headers — user, project, cost center — without touching application logic.
Rules managed in-app
Auto-tagging rules evaluate off the hot path: match on model, size, or existing tags, and stamp the taxonomy your finance team actually wants.
Signals stamped automatically
Purpose, topic, template fingerprints, tool and agent activity, retries — enriched by the pipeline after the response returns. Zero setup.
Performance & reliability
- under 5ms added latency at p99 on the hot path
- responses stream back untouched — token for token
- analytics run off the hot path and can never block a request
See what one swapped URL turns into.
See the live demo — no signupA real product on a rich synthetic organization. No signup, no call, no email gate.