Sources & coverage

Every place your AI coding evidence comes from.

Ottto brings supported local agents, connected accounts, and cloud billing into one source-labeled picture—without sitting in the model-call path.

Start free → Download macOS

Seven current source paths · no request-path proxy · no raw prompts or code in local-enriched uploads

Illustrative reconciliation. Source, cost basis, and freshness stay attached in the product.
Local coding agents

Where the work actually runs.

Point, focus, or tap to inspect with Oto; press a card to keep its details open. Local sources describe observed usage—not an invoice—and content stays out of local-enriched uploads.

Accounts & cloud billing

The reported side of the ledger.

Connected sources stay intentionally narrow: Cursor Team API, eligible Anthropic analytics, Bedrock through Cost Explorer, and Vertex through one selected BigQuery billing export.

Why this gets hard

One tool name can hide several evidence paths. Local observations, account analytics, subscription limits, calculated rates, cloud billing, and credits do not arrive with the same timing or cost basis. Ottto keeps those differences visible while bringing the supported paths together.

Source taxonomy

Five collection methods, each with a clear boundary.

The method defines what the evidence can support—and what it cannot.

  1. 01 Local baseline

    Local enriched

    Reads aggregate usage from supported local session files, adding machine, repo, model, token, and timing context when available.

  2. 02 Opt-in detail

    Live telemetry

    Supported agents can send opt-in liveness and per-turn events. Live data adds detail; local files remain the accounting source.

  3. 03 Connected accounts

    Integration connectors

    Scoped account and admin APIs import usage and team analytics when a local-session path is not available.

  4. 04 Billing actuals

    Cloud billing

    Read-only billing connections import provider-reported cost, credits, and whatever usage detail the billing source exposes.

  5. 05 Derived evidence

    Calculated by Ottto

    Observed usage is priced with Ottto’s catalog. Estimates, normalized comparisons, and coverage assessments stay labeled as derived.

Evidence reconciliation

Keep actuals, reports, and estimates separate.

Where sources overlap, Ottto compares observations and surfaces mismatches without silently changing a primary total’s basis.

  1. 1
    Collect

    Read local rollups, connector API rows, or cloud billing.

  2. 2
    Normalize

    Align source, model, account, machine, and time.

  3. 3
    Label

    Attach cost basis, freshness, gaps, and mismatches.

⌂

Local-enriched uploads

Include aggregate usage and available model or timing context. Exclude raw prompts, responses, code, tool output, local paths, and credentials.

↗

Connected-account access

Uses scoped API or admin permissions and fields. Cursor and Anthropic connections do not collect local transcripts.

◫

Cloud-billing access

Uses AWS Cost Explorer for Bedrock and a selected BigQuery billing export for Vertex. Both are read-only billing paths, not prompt or request proxies.

Start with one source

Connect the sources you already use.

Install the Mac app for supported local agents. Add Cursor Team API, Anthropic, Bedrock Cost Explorer, or a selected Vertex BigQuery billing export from Apps.

Start free → Download the Mac app