Design Stage

AESOP Platform Options

Seven platforms, one architecture recommendation. The Architecture Agent scores each platform against the PRD — evidence-based, no vendor default.

      flowchart TD
        PRD["PRD Compiled"]
        ARCHITECT["Architecture Agent"]
        SCORE["Score Every Platform"]
        RANK["Ranked Recommendation"]
        USER{"User Selects Platform"}
        BUILD["Build Pipeline"]

        PRD --> ARCHITECT
        ARCHITECT --> SCORE
        SCORE --> RANK
        RANK --> USER
        USER --> BUILD

        classDef input fill:rgba(180,83,9,0.08),stroke:#b45309,stroke-width:2px
        classDef agent fill:rgba(74,144,196,0.08),stroke:#4a90c4,stroke-width:2px
        classDef output fill:rgba(15,118,110,0.06),stroke:#0f766e,stroke-width:2px
        classDef decision fill:rgba(124,58,237,0.08),stroke:#7c3aed,stroke-width:2px,stroke-dasharray:5 3

        class PRD input
        class ARCHITECT agent
        class SCORE output
        class RANK output
        class BUILD output
        class USER decision

Enterprise Platforms

Glean

Enterprise knowledge retrieval — connector-based data access, workflow automation within the Glean ecosystem.

Output: agent-config.json + architecture guide + deploy instructions + test scenarios

Constraint: Glean connector required for each data source. Agents are stateless — no persistent caching.

Monday.com

Work OS — boards, automations, and AI agents. Visual workflow management with 24+ column types, native integrations, and MCP programmability.

Output: Board schemas + automation configs + GraphQL mutations + agent JSON

Constraint: Board-first data model. Rate limits by plan tier (1K-25K calls/day).

API / Agent Platforms

Claude API

Maximum reasoning capability — complex analysis, tool use, multi-step agentic workflows, extended thinking. The Anthropic SDK with full control over every parameter.

Output: System prompt + tool definitions + orchestration code + evaluation harness

Constraint: API key required. No built-in UI — you bring the interface.

OpenAI Assistants

Built-in code interpreter, file search (RAG), and function calling within the OpenAI ecosystem. Managed threads and persistent conversations.

Output: Assistant config + function schemas + thread management code

Constraint: Locked into OpenAI ecosystem. Usage-based pricing scales with complexity.

Generic / Platform-agnostic

Portable system prompt — platform TBD, simple Q&A, or proof of concept. Build the agent first, decide where it runs later.

Output: Standalone system prompt + evaluation criteria + integration guide

Constraint: No deployment target specified. The prompt is portable — implementation is on you.

Deployment Targets

Full-Stack Cloud

Production-grade MVP stack: Next.js on Vercel, FastAPI on Fly.io, Supabase for auth and database. Dual LLM support — DeepSeek for prototyping, Anthropic for demos. Live URL, shareable, credible.

Output: main.py + Next.js app + schema.sql + fly.toml + Dockerfile + 6 deploy prompts + DEPLOY.md orchestrator

Constraint: Three cloud accounts required (Vercel, Fly.io, Supabase). First deploy ~20 min.

Local

Self-contained web app that runs on your machine. FastAPI + SQLite + browser UI. Zero cloud dependencies — ideal for demos and prototypes.

Output: main.py + Jinja2 templates + Dockerfile + docker-compose.yml

Constraint: Single-machine only. No cloud services as hard dependency — runs entirely local.

The Architecture Agent scores every platform against the PRD using each platform's weighted scoring rubric. The recommendation is evidence-based — no vendor default, no lock-in. Platforms are ranked 1 through 7 with rationale, best-fit signals, and watch-out-for warnings. The user always has the final choice. Select a platform, then the build pipeline takes over.