[bmdpat]

§ 001 / CURSOR PLUGIN

Agent Architect.

Scope and cost-estimate any AI agent build in seconds. A Cursor plugin that turns a plain English description into a tier, cost breakdown, architecture, timeline, and the top three risks for that specific agent.

No API key. No setup. Free tier: 5 scopes per month.

§ 002 / INSTALL

1. Open Cursor → Marketplace → search Agent Architect.
2. Click Install.
3. In Cursor chat, describe an agent and ask for a scope. Or type /scope-agent.

Open source · github.com/bmdhodl/agent-architect

§ 003 / WHAT YOU TYPE

  • Read my unread Gmail every morning, draft replies for routine ones, post a Slack summary.
  • Multi-agent customer support handling tier-1 tickets with CRM updates and analytics.
  • Daily digest agent that reads my 12 RSS feeds and emails me a summary.

§ 004 / WHAT COMES BACK

Tier: DIY / Startup / Growth / Enterprise
Cost: API $/mo · Infra $/mo · Dev hours · Total build
Architecture: Pattern · Memory · Tools · Key decisions
Timeline: Weeks-to-launch range
Risks: Top 3, specific to this agent

§ 005 / TIERS

  1. DIY$500–$2K · 1–2 weeks

    Single-purpose, minimal state, weekend build.

  2. Startup$2K–$15K · 2–6 weeks

    Multi-step, tool use, basic memory.

  3. Growth$15K–$50K · 6–16 weeks

    Multi-agent, complex state, production reliability.

  4. Enterprise$50K–$150K · 16+ weeks

    Mission-critical, multi-system, compliance.

§ 006 / PRICING

Free tier: 5 scopes per calendar month
Storage: Local at ~/.agent-architect/usage.json
Reset: 1st of every month
API key: Not needed. Cursor's agent does the work.

§ 007 / COST RUNAWAY IS THE #1 RISK

Almost every scope comes back with cost runaway or tool-call loops in the top 3 risks. That is a runtime problem, not a scoping problem. AgentGuard is a Python SDK that enforces budget, loop, timeout, and rate limits on any agent. Drop it in, set the rails, ship to production without paging at 2am.

Install: pip install agentguard47

See AgentGuard

MIT licensed · Open source

← back

Get The One-Person Holdco (free PDF)

How one human plus twenty-two AI agents runs a seven-pillar portfolio with no employees.