Protolap

Receipts for AI developers.

Every AI call your product makes — itemized, timestamped, and on the record. One dashboard with the receipts for every model, every feature, and every user.

Start logging free → See how it works
Works across OpenAI Anthropic Google Azure + any provider
Live ingest protolap.com/logs

Observability built for teams
shipping AI in production.

From a single ingest event, Protolap reconstructs the full picture of your AI usage — what it costs, how fast it is, where it breaks, and who’s driving it.

01

Cost tracking

Real-time and historical spend broken down by provider, model and end user. Auto-estimated when you don’t send a cost, exact when you do.

02

Usage analytics

Token consumption trends, request volume and latency distributions — see which models earn their keep and which quietly bloat the bill.

03

Error monitoring

Error rates and failure patterns across every provider and model — catch a 429 storm before your users ever do.

04

Per-user breakdown

Attribute cost and tokens to each app_user_id with no extra instrumentation. Power usage-based billing and abuse detection.

05

Anomaly detection

Automatic alerts the moment daily cost spikes beyond your normal threshold — know why last Tuesday was expensive, instantly.

Up and running
in three steps.

No SDK to install, no credentials to share. If you can send an HTTP request, you can have full AI observability.

01

Prompt

Keep calling OpenAI, Anthropic, Google, or any other provider exactly as you do today. No changes to your stack, no new dependencies, no credentials to hand over.

02

Send

After each AI call, forward a single HTTP POST with the metadata — model, tokens, latency, cost, user ID. That’s it. No SDK, no prompt storage, no agents to babysit.

03

Observe

Protolap normalizes every event across providers into one schema and surfaces it in real time — spend, usage trends, error rates, and per-user breakdowns, all in one dashboard.

2-step integration.

Point your AI calls at Protolap in minutes — fetch your schema once, then send one event per call.

Step 01 · once per session GET
https://protolap.com/api/ingest/schema

Fetch your workspace event schema. It’s the source of truth — which fields to collect, their types, and which are required.

Step 02 · per AI call POST
https://protolap.com/api/ingest/event

Build one event from the schema and send it. Fire-and-forget — logging never blocks or breaks your app.

→
AUTH Authorization: Bearer pk_your_key_here Send this header with both requests.

Built for
multi-agent workflows.

When an orchestrator calls three models to answer one question, you need to see the whole chain — not just the last hop.

01

Full pipeline visibility

Log every model call in your agent chain. Pinpoint exactly which step is slow, expensive, or failing — without guessing.

02

Session-level tracing

Group related calls with a shared session or correlation ID. Reconstruct the full sequence of agent calls for any user, any run.

03

True cost attribution

Multi-step pipelines multiply your token spend. Attribute cost to each agent role so you know what each step actually costs to run.

Two tiers. No surprises.

Both plans give you the full Protolap ingest pipeline, analytics, and log explorer.

Basic
$8.99 /mo

Everything you need to monitor and understand your AI costs.

  • —AI call logging & batch ingest
  • —Log Explorer with filters & search
  • —Analytics — cost, tokens, latency
  • —Per-user AI analytics
  • —OpenAI & Anthropic proxy mode
  • —Cost estimation & anomaly alerts
  • —Custom metadata fields
  • —Team member access
  • —Weekly email digest
Get started →
Enterprise
$14.99 /mo

Full platform access plus collaboration and reporting tools.

  • —Everything in Basic
  • —Custom metadata fields up to 20
  • —Delegated team access
  • —Weekly email digest
Get started →

Get the receipts.

Turn every AI call into a structured record — so you can answer any question, instantly.

Start logging free → See pricing