Model choice is yours; two rules are not.
Assign a model to a role the way you would assign a specialist to a job: a long-context model for Blueprint architecture, a fast one for routine changes, a stronger reasoner for critique. Routing is configuration, so it moves when the model market moves and your workflows stay as they were. Two constraints hold whatever you set — the model behind a proposal cannot be the model that reviews it, and workflows marked critical must run across more than one model provider. Every run records which model did which job, so any output traces back to its source.
A single model is a single failure mode
One model's blind spot becomes the system's blind spot when the same weights draft a change and then read it back. Agreement between a model and itself looks like verification and demonstrates only consistency. Two independent reads fail together far less often than one read fails alone, which is why critical work is spread across families.
Routing rules live in configuration
Each role carries a model assignment; each workflow carries a policy profile. At the start of a run the platform resolves the assignment, tests it against the self-review and diversity rules, and refuses the run when a rule cannot be met. None of this rests on prompt wording, which any model can be argued out of.
Swap models without rewriting work
A better model arrives; you change the assignment on the role and the next run uses it. Blueprints, capabilities, policies and generators are untouched, because none of them names a model. Cost and latency become tuning decisions rather than migrations.
In this capability
- Model assigned per role
- Self-review blocked at routing
- Diversity required on critical work
- Model recorded on every run
AI Workforce
The rest of AI Workforce
Agent Roles
Portfolio Strategist through Release Manager: every role has a written scope, defined inputs and work it is not allowed to do.
Separation of Duties
An agent never approves its own output. Proposal and review are separate roles, held by separate agents, every time.
Agent Permissions
An agent holds a scope, not a key to everything: what it may propose, what it may change, what it must escalate.
Decisions, Not Transcripts
The interface shows the decision, its reasoning, its impact and the options — not a scroll of agent conversation.
Stop managing prompts. Start managing software strategy.
Connect your repositories. Let Foundariq understand what you have already built. Then tell it what your portfolio should build next.
Read-only to start · Your code stays in your GitHub · [email protected]