Creative AI agent discoveryFrom Sidekick Studios ↗
sidekickSTOREHelp me choose
The Sidekick Store

Evidence before
endorsements.

An open standard for evaluating AI against real marketing and creative work.

A good creative workflow needs more than an impressive demo. Our aim is to publish testing that a marketing team can understand and repeat.

Current status: research only. The launch catalogue uses provider information checked on 27 October 2026. No listing has completed Sidekick’s hands-on test protocol. We do not publish independent scores, security approvals or performance claims without evidence.

Creative agents only

To qualify, a product must perform a creative task and have provider documentation of agent-led planning, tool selection or multi-step execution. We list the specific agent experience rather than treating every feature of a creative suite as an agent. General research, sales, CRM, automation-building platforms and single-step generators are outside this directory.

Three evidence levels

  1. Research only: a dated summary of public product information. No independent endorsement.
  2. Hands-on tested: a documented brief, test runs, outputs, costs and limitations. This label is reserved for completed tests.
  3. Scored: a completed test set reviewed against the published rubric and security gate. Scores will include the test date and product version where available.

The proposed Sidekick Score

Each dimension is scored from 0 to 5, then weighted. The final result is reported out of 100. We will publish the individual dimension scores alongside the overall result.

Before a score can be published

  1. Record the product, version or plan, date, brief, input assets, configuration and success criteria.
  2. Run the same core task at least five times and record failures as well as successes.
  3. Have two reviewers assess creative output against the brief and reconcile significant differences.
  4. Record actual usage costs, setup time and human editing time.
  5. Apply a security gate: unresolved material concerns about permissions, data handling or required approvals block a recommendation, regardless of creative quality.
  6. Publish sample outputs only where we have permission, with limitations, evidence links and the next review date.

Commercial independence

A free listing, affiliate relationship or future sponsorship cannot buy a score. Any paid placement will be labelled separately and excluded from score-based ordering.

Read the editorial policy →