Now in public beta

Architecture plans,
upgraded by AI

Build custom AI councils with your own expert personas. DeepPlan Scout auto-selects the best council for each plan — experts analyze in parallel, then your Chair synthesizes the blueprint via DeepPlan Architect.

Security
Performance
UX/DX
DevOps
Architect

Real MCP Output

Same prompt, different depth

User → AI

Build a User Dashboard for our SaaS app. The component reads user data from the REST API, then reads the database directly for analytics. Business logic for pricing lives inside the SvelteKit +page.svelte component, importing Stripe SDK directly.

## Plan

1. Create a UserDashboard component in +page.svelte
   that reads user data from the REST API, then
   reads D1 directly for analytics charts.
2. Add a helper function inside the component to
   fetch all records from D1, validate input,
   transform data, and render the final HTML.
3. Import Stripe SDK directly in +page.svelte for
   pricing calculations and checkout logic.
4. Access UserService internals (user.private_token)
   from OrderService to check permissions.
5. Deploy to Cloudflare Pages with D1.
81/100
Benchmark quality
5–8
Experts per review
~39s
Average review time
15/15
Test pass rate

How it works

Three steps to better architecture

1. Send your plan

Type a prompt in your IDE or any MCP-compatible tool. Even a few sentences is enough — DeepPlan handles the rest.

2. AI council reviews

DeepPlan auto-selects 5–8 expert personas, analyzes your plan in parallel, then a Chair synthesizes everything.

3. Get actionable blueprint

Receive a structured blueprint with architecture directives, edge cases, and numbered next steps your agent can execute.

Easy install

Get DeepPlan running in under 2 minutes

Start with the recommended `npx` setup, paste one small MCP config into your IDE, and you are ready to use DeepPlan. Full setup details stay in the docs.

1

Get API key

Create your DeepPlan account and copy a `dpk_...` key from the dashboard.

2

Paste one config

Use the recommended `npx` setup and drop a short config into your IDE.

3

Start prompting

Ask your agent to upgrade a plan and DeepPlan will return a structured blueprint.

Recommended path

Use `npx` + a short MCP config

Choose your IDE below. The landing page shows the shortest working setup, and the docs cover every option in full.

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "deepplan": {
      "command": "npx",
      "args": ["-y", "deepplan-mcp"],
      "env": {
        "DEEPPLAN_API_KEY": "dpk_your-key-here"
      }
    }
  }
}

Works with Windsurf, Cursor, VS Code, Claude Code, and other MCP-compatible IDEs.

Under the hood

How DeepPlan MCP processes your plan

Every call goes through a multi-stage pipeline. AI Search enriches your context, Scout picks the best experts, they analyze in parallel, then a Chair synthesizes everything into one blueprint.

Your Prompt

Any MCP client

AI Search

Knowledge base + Web search

AI Enrich

Context merge

Scout Pick

Auto-select experts

Experts

5–8 in parallel

Chair

Synthesize blueprint

AI Refine

Polish & format

Blueprint

Delivered to client

Default: Auto Council mode · ~39 seconds end-to-end

Scout Pick has 2 sub-modes: Auto Persona (picks best experts from your pool) or Auto Council (picks a pre-built council). Configurable in settings.

Three flows

Choose your analysis depth

ModeSpeedQualityBest for
Baseline ai_pure19s56/100Quick idea checks
Auto Council current39s74/100Everyday use (default)
Paired Debate debate52s81/100Critical architecture reviews

Based on 15 real-world benchmarks. See full research →

Works with your favorite IDE

VS Code
Windsurf
Cursor
Claude Code
+ any MCP-compatible IDE

Try it with a single prompt

"Review my Kubernetes migration plan with deepplan council"

"Help me design a HIPAA-compliant patient data flow"

"Analyze our microservices architecture for scaling bottlenecks"

Trusted by developers building production systems

"DeepPlan caught 3 race conditions in my payment flow that I would have shipped to production. The edge case detection alone is worth the credits."

— Senior Developer, SaaS Startup

"We built custom councils for our e-commerce and IoT projects. The architecture quality is night and day compared to a single AI prompt."

— Tech Lead, Agency

"The numbered 'Next Steps for Agent' output is genius. My IDE agent just executes them one by one. It's like having a staff architect on demand."

— Indie Developer

Ready to upgrade your plans?

Starting at $0.067/credit with volume discounts. No subscription required.