Adopt / By Team Size
Adoption by team size
The framework works for solo developers and for enterprise platforms. Your team size determines where you start, not where you end up.
Solo developer
Setup: 1-2 hoursYou are the team. You configure the agentic layer, agents execute it. No coordination overhead, no approval workflows. Just you and your workflow.
Recommended workflow
Start with four phases: Plan, Build, Test, Deploy. Skip Review, Document, and Monitor for now. You're the reviewer. Add phases when you feel the need, not before.
Lightweight gates
The test-pass quality gate from Getting Started is sufficient. One gate, one criteria: do the tests pass? Add coverage thresholds and lint checks when you're ready for Level 2.
Focus area
Prompt templates that encode YOUR coding standards. Your agent should produce code that looks like you wrote it. Spend your time on constraints, conventions, and architectural patterns specific to your project.
Maturity target
L2-L3 for most work. L4 for routine maintenance tasks (dependency updates, simple bug fixes). You'll get there faster than teams because there's no coordination overhead.
Here's a minimal workflow.yaml for solo developers:
phases: [plan, build, test, deploy]
loops:
test_retry: { max: 3, on_exhaust: escalate }
gates:
test: { all_pass: true }
Start here: complete the Getting Started guide. It's designed for solo developers and takes less than an hour. Everything else builds from that foundation.
Small team (2-10 engineers)
Setup: 1-2 daysYour agentic layer becomes a shared artifact. Store it in your repo and version it like code. The hard part at this size is consistency: every engineer's agent needs to produce work that meets the same standard.
Full workflow
Use all seven phases of the workflow: Plan, Build, Test, Review, Document, Deploy, Monitor. The Review phase is no longer optional. It's how you keep consistency across the team. Document catches what humans would forget.
Standard gates
Add 2-3 quality gates: test-pass, review-pass, and a coverage threshold. These gates enforce team standards without putting a human reviewer on every change.
Template ownership
Template changes go through PR review. Senior engineers maintain templates; all engineers use them. This is the core of teams collaborating on the layer, not the code.
Maturity target
L2-L3 for most work. L4 for routine tasks with well-tested templates. Graduate levels per issue type: patches can be L4 while features stay at L3.
Division of labor
| Senior engineers | All engineers |
|---|---|
| Write and maintain prompt templates | Use templates for daily work |
| Configure quality gates | Run workflows against issues |
| Tune loop limits and escalation rules | Report template issues |
| Review template change PRs | Propose improvements |
Start here: have one senior engineer complete Getting Started, then share the agentic layer directory with the team. Iterate on templates together. Read the Customization Guide for adding team conventions.
Enterprise (10+ engineers)
Setup: 1-2 weeksYour agentic layer becomes a platform. A dedicated team maintains the base layer (shared templates, standard gates, workflow defaults). Individual teams extend it with overrides for their domains.
Base layer + team overrides
Maintain a shared base agentic layer with organization-wide defaults. Each team adds overrides for their stack, conventions, and domain. Team templates replace base templates for the same phase; team gates are additive. See the Customization Guide for the directory pattern.
Risk-tiered gates
Not every change needs the same scrutiny. Configure quality gates by risk profile: strict gates for critical services (security review, architecture review), standard gates for business logic, lighter gates for internal tools and documentation.
Audit and governance
Every workflow run produces artifacts: the plan, build output, test results, review verdicts. These are your audit trail. Template changes require approval from the platform team. Gate results are logged for compliance reporting.
Maturity target
L2-L4 tiered by risk. Critical paths stay at L2-L3. Routine maintenance across trusted repositories can reach L4. Use the Monitor phase to track KPIs across teams and identify where autonomy is safe to increase.
Key practices
Base layer + team overrides directory pattern
organization-level defaults that every team inherits, with per-team directories for specialization. Team templates replace base templates for the same phase. Team gates stack on top of base gates.
Risk-based gate configuration
three profiles (critical, standard, low-risk), each with different gate strictness, loop limits, and escalation paths. Map repositories or services to risk profiles.
KPI monitoring across teams
track workflow success rate, average loop count, escalation rate, and time-to-PR per team. Use these metrics to identify which teams are ready to advance maturity levels and which templates need improvement.
Start here: have the platform team complete Getting Started to establish the base layer. Then pilot with one team for 2 weeks. Use that team's feedback to refine templates before rolling out to the organization. Read Customization for the team overrides pattern.
Side-by-side
| Dimension | Solo | Small team | Enterprise |
|---|---|---|---|
| Setup time | 1-2 hours | 1-2 days | 1-2 weeks |
| Workflow phases | 4 phases | All 7 | All 7 + custom |
| Gate complexity | 1 gate | 2-3 gates | Risk-tiered |
| Template ownership | You | Senior engineers | Platform team |
| Maturity target | L2-L3 | L2-L3 | L2-L4 tiered by risk |
| Coordination | None | Template PRs | Platform governance |