Skip to content

Activity methodology

From local agent records to a durable daily activity history.

Agentprint discovers supported coding-agent records on your machine, converts recognized numeric metadata into a shared format, and syncs idempotent batches into your profile.

Agentprint measures activity volume and continuity. Token totals are not a score for code quality, output, or developer productivity.

Start with Agentprint See how it works
4 agentsSupported activity sources
IdempotentRepeated syncs do not inflate totals
DailyActivity is normalized by date
The useful outcome

A comparable history without pretending every tool emits the same files.

Each adapter understands its own local source. The normalized record keeps only the fields Agentprint can compare responsibly across supported agents.

  1. 1

    Discover supported local sources

    The collector checks known locations for Claude Code, Codex, OpenCode, and Kimi Code and reports what it finds.

    agentprint sources
  2. 2

    Normalize recognized numeric fields

    Dates, token counts, coding-tool identity, version, provider, and model are mapped into a strict shared record. Unknown or content-bearing fields do not enter the batch.

  3. 3

    Sync an idempotent batch

    Each batch carries stable identity so retries and later scans update the history without double-counting previously accepted activity.

    agentprint sync
  4. 4

    Build the visible history

    Accepted daily records power the activity field, lifetime totals, coding-tool mix, model ranking, and streak calculations.

Built into the workflow

Useful by design, clear about the boundary.

Activity is descriptive

A busy day can reflect exploration, debugging, or repetition. Agentprint shows the trace without converting it into a performance grade.

Adapters stay source-aware

Claude Code, Codex, OpenCode, and Kimi Code keep separate readers because their local record formats and capabilities differ.

Retries remain safe

Stable batch and record identities let the collector retry offline work without multiplying previously accepted totals.

Frequently asked questions

Does a larger token total mean better work?

No. Token volume is an activity measure, not a quality or productivity score. It can be influenced by task complexity, model behavior, context size, and workflow style.

Can Agentprint count the same activity twice?

The sync protocol is designed to be idempotent: records and batches use stable identities so resubmitting accepted work does not inflate the activity history.

Why are coding-agent adapters separate?

Each supported agent stores local activity differently. Separate readers preserve those differences while emitting the same strict numeric contract.

Does this methodology include transcript content?

No. Automatic activity collection and explicit session sharing are separate systems with separate contracts and endpoints.

Keep exploring.

Follow the product from the question you have now to the exact workflow that answers it.

Collection boundarySee the accepted and excluded fields.
Supported integrationsInspect each source-specific workflow.
AI coding activity trackerSee what the normalized history is useful for.