Agents are the components teams are built from — each with its own role, model, platform, and track record. 385 components on record. Tiers are computed from evidence, not self-assigned.
HAMY
Test design assessment
Test execution
Shopify
Reads the reference code and writes test cases for every checkpoint.
Alexey Grigorev
Checks acceptance criteria and reports test evidence.
Oleh Veheria
Checks whether tests cover the behavior change.
Allie (AI with ALLIE)
Takes notes for the team.
Zembl
The first agent, live since January 2025; researches leads, makes contact and collects their energy bills.
Minimal
Executes the planner’s tool plan.
Mira Systems
Context-aware localization across 12 languages.
imkisyou
English adaptation
Redazione Zero
Checks that every statement is supported by the source it cites and can stop a piece from being published.
Panasonic Avionics Corporation
Detects service degradation and anomalies.
Uber (Developer Platform team)
Builds, executes, and measures coverage for Uber AutoCover.
Generates new unit test cases in Uber AutoCover.
Environment prep and business-case identification for Uber AutoCover.
Central coordinator dispatching Uber Validator sub-agents.
Deterministic static-analysis sub-agent inside Uber Validator.
Curated best-practice review sub-agent inside Uber Validator.
Searches the web in 11 languages and gathers facts together with their sources into dossiers for the writer.
Fastweb + Vodafone
Each one resolves a single category of customer need, for example billing, roaming, sales, active offers or cost control.
PGA TOUR
Review drafts for quality and for compliance with the TOUR's brand rules.
Next.js
Checks issues against the canary release.
NVIDIA Research (Voyager)
Open-ended Minecraft agent — 3.3x items; tech tree up to 15.3x faster (wooden tool level) vs prior SOTA.
Cursor
Searches existing code for security flaws.
Jenny Ouyang
Builds and deploys websites; she and her business partner both work with it in a Telegram group.
Bertelsmann
Tracks outside trends and commentary to add context from beyond the content Bertelsmann owns.
Microsoft Research
Web browsing specialist in the Magentic-One system.
Antonio Leiva
Once a week it reviews how his posts performed over the past 15 days and suggests new ideas.
Anthropic
Specialist LLM executing delegated subtasks from an orchestrator.
Andon Labs
Handles the in-store voice kiosk and the store's phone line.
Japanese drafting
Write the content itself, in a range of formats and styles.
Hyuk Jun Kim
Sub-agents that write code changes and research in isolated cloud copies of the repo, then hand over bundles. They never deploy.
monday.com
Writes updates from findings.
Chooses the stories, ranks them and writes and signs the daily editorial, which comes last in the run.