CasperWasp
Get started
How it works · Platform

Per-call audit trail

Every AI call logs app, tool, model, tokens consumed, and outcome so organisations can reconstruct usage without guesswork.

Platform

Overview

Opaque AI products tell you that you “used a lot” this month. CasperWasp records each call: which studio and tool ran, which model served it, how many tokens were consumed, how long it took, and whether it succeeded. That ledger is how a marketplace stays transparent across Document Studio and Design Studio.

The audit trail answers operational questions. Which team burned tokens on logo exploration? Did PDF chat spike during diligence week? Did failed provider calls quietly look like usage — or correctly show zero tokens? Without per-call data, shared usage becomes a feeling instead of a system.

Importantly, the trail is about AI usage, not about replacing document version history or approval status. Versions explain how text changed. Approvals explain who blessed a pack. The audit trail explains which model calls happened and how many tokens they consumed. Mature organisations use all three.

There is no token calculator substituting for the ledger, and no pre-run estimate that pretends to know your prompt’s cost before you run it. You work, the system logs, you review. Editing and export never appear as token lines because they are free.

In depth

Understanding Per-call audit trail

How the feature works in practice inside the CasperWasp marketplace — and how it connects to the rest of your work.

What each log line should tell you

A useful entry names the app (Document Studio, PDF Intelligence, Design Studio), the tool (rewrite, summarise, logo generation, and so on), the model, token counts, duration, and outcome. That is enough to attribute usage without reading prompt contents in a finance meeting.

Outcome matters because failed calls use zero tokens. If a run errors, the trail should still exist for debugging and support — but it should not punish the wallet. Success lines carry the consumption that shared usage meters.

Aggregations by studio and tool help admins set norms. If image generation dominates a month, maybe the team needs stronger creative briefs. If full-document generate dominates, maybe narrower rewrite tools would suffice.

Export the story when finance needs reconciliation. A month reconstructed from the ledger beats a month reconstructed from memory and screenshots.

How the trail supports roles and spending gates

Editors generate; owners and admins interpret and act. The audit trail gives admins evidence to decide whether to buy a token pack, upgrade a tier, or coach the team on heavier tools.

Viewers may never create usage, but admins still need clarity when explaining organisation consumption. Role design keeps spending gated while the ledger stays readable for people who hold the wallet.

If a contractor’s work spikes tokens, the trail helps you attribute the spike to tools and apps — then decide whether the spike was valuable. Shared usage without attribution becomes political; with attribution it becomes manageable.

Security-minded teams should remember: AI prompts may reach third-party model providers. The audit trail logs usage metadata for operations; it is not a claim of SOC 2 or a substitute for reading the security page.

Pairing usage logs with document collaboration history

When reconstructing “what happened to this MSA,” open version history for text, approval status for gates, and the audit trail for which AI tools touched the draft. Those layers answer different auditors.

A bad AI expand that was restored still appears as tokens consumed for the successful run. That is correct: the model work happened. Restore is free editing afterward. The ledger and the version pin together tell the honest story.

Design explorations are similar. Multiple logo generations may be discarded visually but still consume tokens on success. The trail helps creative leads teach “brief better, generate less thrash.”

Because CasperWasp is a marketplace, one trail covers writing and design. You do not reconcile three vendor portals to understand last month.

How it works

Step by step

A practical walkthrough of per-call audit trail in the CasperWasp marketplace.

  1. 01

    Run AI tools as part of normal work

    Write, analyse PDFs, or generate design assets. Successful calls consume tokens; editing remains free.

  2. 02

    Let the platform write the log automatically

    Each call records app, tool, model, tokens, timing, and outcome without a manual timesheet.

  3. 03

    Open the usage audit view

    Admins and owners review recent calls and aggregates for the organisation.

  4. 04

    Filter by studio or tool

    Separate document rewriting from image generation to understand the mix that drove tokens consumed.

  5. 05

    Verify failed calls show zero tokens

    Confirm error outcomes did not drain the pool. Escalate if something looks wrong.

  6. 06

    Coach the team with evidence

    Use real call patterns to encourage version checkpoints and narrower tools where appropriate.

  7. 07

    Decide on packs or tier changes

    Owners/admins act on the ledger: top up allowance or unlock higher model tiers if needed.

  8. 08

    Export or summarise for finance when required

    Reconcile the month from per-call truth rather than from estimates that were never shown pre-run.

When

When to use this

  • Monthly finance reconciliation of AI tokens consumed
  • Investigating a sudden spike in organisation usage
  • Coaching teams that overuse full regenerations
  • Separating document vs design consumption in shared usage
  • Confirming failed calls did not consume tokens
  • Preparing for a plan tier discussion with evidence
  • Agency producers explaining usage to clients or internal finance
  • Any audit question that starts with “which tool did that?”
Who

Who it is for

  • Owners and admins responsible for plans and token packs
  • Finance partners reconciling marketplace AI usage
  • Ops leaders setting responsible AI norms
  • Agency producers attributing usage across projects
  • Security-conscious admins reviewing operational metadata practices
  • Team leads investigating tool-level spikes
  • Anyone who wants tokens consumed to be inspectable
Examples

Real workflows

Concrete jobs teams run with this feature — not abstract capability lists.

Finance asks why tokens jumped: the trail shows Design Studio logo runs during a rebrand week, not Document Studio drafting.
An editor reports a failed rewrite; the admin confirms the call outcome and zero tokens, then retries after a provider blip.
A legal ops lead correlates a diligence week’s PDF chat spike with audit lines by tool and day.
An agency producer exports usage to show a client which AI-assisted tasks ran on their pack.
An admin denies a blanket upgrade request after seeing that most calls are light rewrite tools already covered by the current tier.
A founder restores a bad AI expand, notes the successful consume in the trail, and teaches the team to checkpoint first.
Included

What you get

  • Per-call logging across marketplace studios
  • App, tool, model, and token fields for attribution
  • Outcome visibility including failed calls at zero tokens
  • Aggregations that support shared usage reviews
  • Evidence for pack purchases and tier changes
  • Separation from document version and approval history
  • A finance-readable story in tokens consumed
  • No reliance on pre-run cost estimates
  • No token calculator pretending to replace the ledger
  • Organisation-scoped visibility for admins and owners
  • Support for coaching healthier generation habits
  • Transparency that matches the marketplace positioning
Tips

Do it well

Review usage weekly during campaigns, monthly otherwise.
Teach “failed = zero tokens” so the team trusts the system.
Pair trail spikes with version history when reconstructing incidents.
Attribute before you blame — check which tool and app moved.
Export when finance needs a durable artifact.
Use the trail to choose between packs (allowance) and upgrades (models).
Keep spending rights limited while making usage visible to decision-makers.
Read security documentation for provider and prompt handling context.
Pitfalls

Common mistakes to avoid

The shortcuts that waste time or produce weak deliverables.

  • Treating the audit trail as a document approval log
  • Expecting pre-run estimates instead of post-run lines
  • Ignoring studio breakdowns and managing only a single total
  • Assuming failed calls consumed tokens without checking outcome
  • Giving spending power to everyone who can view usage
  • Reconciling from memory after a busy month
  • Confusing free exports with metered AI calls
FAQ

Common questions

Related

More how-it-works pages

Platform

Shared AI usage

One subscription meters AI tokens across Document Studio, PDF Intelligence, and Design Studio — editing and export stay free.

Platform

Organisations and roles

Owner, admin, editor, and viewer roles across shared workspaces — with spending gated to owners and admins.

Platform

Tier-gated model catalogue

Plans unlock text and image models from Basic through Elite so each tool runs on models your tier allows.

Collaboration

Version history

Numbered saves with compare and one-click restore so risky edits and AI rewrites never erase your last good draft.

Document Studio

Document Studio

Draft, refine, collaborate on, and export business documents with AI beside the cursor inside CasperWasp.

See pricing

Part of the CasperWasp marketplace — documents and design under one subscription.

See pricingAll features

Try the CasperWasp marketplace

Create an account and use Document Studio and Design Studio from one subscription — write documents, analyse PDFs, and generate brand assets with transparent AI usage.

Create an accountSee pricing

Plans from ₹300/month · Cancel anytime