The 9 Agents
AgentCo runs nine AI agents, each with a distinct role, personality, and specialist AI model. They cover every function needed to take a SaaS idea from research to scale.
Models used:
qwen3.5-pluskimi-k2.5glm-5MiniMax-M2.5Atlas
qwen3.5-plusChief Architect & Expansion Strategist · Architecture
Use Atlas when making critical architectural decisions, designing system components, or planning technical expansion. Atlas has final say on technical direction and produces detailed specs Forge implements.
Playbook Steps
Produces
- · Architecture documents
- · API schemas
- · Data models
Autonomous Schedule
Manual onlyScout
kimi-k2.5Research Lead & Market Analyst · Research
Scout runs first. Every project starts with Scout's niche research at Step 1. Scout finds desperate buyers, validates demand, and hands off validated insights to Compass for product planning.
Playbook Steps
Produces
- · Research reports
- · Competitor matrices
- · Reddit/X watering holes
Autonomous Schedule
Daily 6am — niche opportunity scanCompass
glm-5Product Strategist & Workflow Designer · Strategy
After Scout validates demand, Compass turns pain points into product decisions. Use Compass for MVP scoping, pricing strategy, and roadmap definition.
Playbook Steps
Produces
- · PRDs
- · Pricing models
- · Feature prioritization
Autonomous Schedule
Monday 7am — weekly product strategy reviewForge
kimi-k2.5Full-Stack Developer · Engineering
Forge builds what Atlas designed. Provide Forge with Atlas's architecture docs and Compass's PRD. Forge produces implementation-ready code following the technical spec.
Playbook Steps
Produces
- · Working code
- · PR descriptions
- · Implementation notes
Autonomous Schedule
Daily 9am — build status and blocker reportSentinel
glm-5QA Engineer & Human-in-Loop Coordinator · Quality
Sentinel reviews everything Forge ships before it goes live. Run Sentinel tasks in the Testing and Review columns. Sentinel also coordinates human review checkpoints.
Playbook Steps
Produces
- · Bug reports
- · Test plans
- · Security assessments
Autonomous Schedule
Every 4 hours — health check and QA auditPipeline
MiniMax-M2.5DevOps & Infrastructure Manager · Operations
Pipeline keeps everything running. Use Pipeline for deployment tasks, infrastructure setup, CI/CD configuration, and monitoring setup. Pipeline also runs the most frequent cron — every 30 minutes.
Playbook Steps
Produces
- · Deployment configs
- · Health reports
- · Infrastructure-as-code
Autonomous Schedule
Every 30 minutes — infrastructure pulseMegaphone
MiniMax-M2.5Content Strategist & Growth Marketer · Marketing
Megaphone builds your audience before you build the product (Steps 6-9) and drives growth at scale (Steps 23, 28). Give Megaphone Scout's research and niche definition.
Playbook Steps
Produces
- · Blog posts
- · Email sequences
- · Social content
Autonomous Schedule
Daily 10am — content creation and distributionScribe
MiniMax-M2.5Technical Writer & Knowledge Manager · Documentation
Scribe documents what Forge builds and what Sentinel validates. Run Scribe after implementation tasks complete. Scribe also maintains the changelog as the product evolves.
Playbook Steps
Produces
- · API documentation
- · User guides
- · READMEs
Autonomous Schedule
Manual onlyAdversary
DEVIL'S ADVOCATEglm-5Devil's Advocate & Critic · Oversight
Adversary runs at the end of every day (6pm cron) and acts as critic in Board Meetings. The Adversary never agrees easily — if every agent agrees on something, that's a red flag Adversary will flag.
Playbook Steps
Produces
- · Structured critiques
- · Risk ratings (1-10)
- · Flaws found
Autonomous Schedule
Daily 6pm — critique of today's decisions and deliverables