decisions/ops/helicone-prompt-tracking-2026-07-06

concept

helicone-setup-2026-07-06.md

type: decision title: "Helicone cloud for prompt/response tracking" description: "Helicone proxy wraps all LLM calls. 100K req/mo free. 1-line URL change per agent." tags: [observability, llm, helicone, tracking] timestamp: 2026-07-06 format_version: okf-v0.1 status: active

Helicone — prompt + response tracking

Decision

Use Helicone cloud (free, 100K req/mo) for LLM observability.

Setup (pending API key)

1. Sign up: https://helicone.ai
2. Get API key from dashboard
3. Add to .env: HELICONE_API_KEY=sk-helicone-...
4. Run: pwsh repos/private/agent-configs/scripts/sync-configs.ps1

How it works

All agents route via Helicone proxy:

• CC: ANTHROPIC_BASE_URL = https://gateway.helicone.ai/anthropic/
• OpenCode/OmniRoute: add Helicone as a proxy layer in front of :20128
• Header: Helicone-Auth: Bearer ${env:HELICONE_API_KEY}

What gets tracked

• Every prompt + response
• Token counts + cost per request
• Latency per provider
• Model routing (which OmniRoute leg won)

Action needed

Run: ! start https://helicone.ai to sign up, then paste key here.