Research notes · checked July 2026Installer documentation
COMPARISONS · CHECKED JULY 20, 2026

Compare operating models, not feature-count theater

This framework examines workflow, data provenance, control, and cost shape. It does not claim current competitor pricing or certification status. Public facts can change; Unknown means the available material did not establish an answer.

Decision dashboard

WORKFLOW

Separate application or agent-native skill?

A conventional platform centralizes administration and reporting. An agent-native skill reduces context switching and makes natural-language tasks portable. Neither architecture is universally superior; evaluate governance and operator behavior.

DATA

Single source or waterfall enrichment?

A single source can simplify provenance and predictable cost. Waterfall enrichment may expand coverage while introducing provider conflicts and variable cost. Ask how every field identifies its origin.

CONTROL

Autonomous sending or human review?

Autonomy can increase throughput. Review gates protect relevance, regional policy, sender reputation, and brand voice. This implementation assumes people approve targeting, messaging, and sending.

COST

Seats, credits, providers, and runtime

Total cost can include user seats, data credits, verification, APIs, infrastructure, and operator time. Compare a representative task rather than a headline price; mark unavailable inputs Unknown.

DimensionSeat-based platformAgent-native skillEvidence to collect
Setup pathWorkspace and user provisioningPackage review, run, configureAdmin steps and permissions
Task modelUI-defined workflowsNatural-language briefReproducibility of criteria
Data disclosureVaries by vendorConfig-dependentField-level provenance
Human reviewVaries by workflowRequired by this methodApproval event and owner
ExportVaries by planRuntime-dependentSchema and audit notes
Price visibilityUnknown unless publicly verifiedInspect package/provider costsDated cost worksheet

Evidence packet to assemble

01

Representative task brief

Use one ICP, one geography, one buyer responsibility, and a fixed result count. Keep it identical across evaluated approaches.

02

Field provenance sheet

Record source, observation date, verification state, conflicts, and unknown values for each material field.

03

Control map

Identify who approves accounts, messages, send domains, exports, credentials, retention, and deletion.

04

Measured cost model

Include seats, credits, data providers, runtime, integration effort, and operator review time.

Comparison protocol

  1. Day 0

    Freeze the brief, definitions, and acceptance criteria.

  2. Day 1

    Run a small sample and record every unknown or conflict.

  3. Day 2

    Have an independent reviewer judge relevance without knowing the workflow.

  4. Day 7

    Reconcile cost, operator time, deliverability indicators, and governance findings.

Run the same brief through your agent and preserve the evidence

npx -y @okki-global/okki-go-taroball