VERIFIED 12-SLOT ARCHITECTURE BOM • PERSONA: DEVSECOPS / CISO / HEALTHCARE & FINTECH COMPLIANCE LEAD
HIPAA & Defense Zero-Egress Air-Gapped Local Stack (0% Cloud Telemetry)
Monthly TCO: $519.00/mo (Saves $877.92/mo) • Cumulative p50 Latency: 65.3 ms • Concurrent RAM: 2014 MB
INTERACTIVE STACK BUILDER
11 / 12 SLOTS CONFIGURED • 1 OMITTED BY DESIGN
Build Your AI Stack in 3 Simple Steps
Pick a starting preset below, toggle only the tools you actually need (skip the rest), and export your ready-to-run config bundle.
Est. Monthly Cost
$519.00/mo
Saves $877.92/mo
RAM / GPU Footprint
2014 MB / 16.0 GB
65.3 ms p50
License & Privacy
VERIFIED CLEAN
0 swap(s) ready
1
Pick a starting template
Start from a ready-made stack for your situation, then change anything. Cost, RAM and license checks update as you go.
[+] Fine-Tune Compliance Mode, Monthly Session Volume, Compute Power & Budget Cap
2
Seat a tool in each slot
Each slot is one layer of your AI stack. Choose one tool per slot, or skip the layers you do not need. Hover the i for what each layer does.
AI Stack Motherboard
● 6/12 SOCKETS SEATED • READY TO SHIP
Export Seated Stack Configs:
Socket 01Core Layer • IDE
AI Coding IDE & EditoriWhat Socket 01 DoesYour primary code editor for inline AI diffs, multi-file edits, and tab autocomplete.Essential for daily interactive coding • Skip only for headless server pipelines.
● Seated
+ Add a tool that is not listed
We cannot check unlisted tools for license or speed. We use your answers for the cost and privacy checks, and list it by name in your export.
Socket 02Core Layer • CLI Agent
Autonomous Terminal & SWE AgentiWhat Socket 02 DoesAutonomous terminal agent that plans, edits across your repo, runs tests, and commits.Best for multi-file refactors & PR automation • Optional if you only use an IDE.
● Seated
+ Add a tool that is not listed
We cannot check unlisted tools for license or speed. We use your answers for the cost and privacy checks, and list it by name in your export.
Socket 03Core Layer • Context OS
Agent Harness & Context OSiWhat Socket 03 DoesCompresses repo context (CLAUDE.md / AST maps) so agents stop re-reading junk files.Recommended for all stacks • Cuts per-turn prompt token waste by 40% to 75%.
● Seated
+ Add a tool that is not listed
We cannot check unlisted tools for license or speed. We use your answers for the cost and privacy checks, and list it by name in your export.
Socket 04Core Layer • Cost Router
LLM Gateway, KV-Cache & FinOps RouteriWhat Socket 04 DoesRoutes simple prompts to fast/cheap models and caches repeated context (65/35 cascade).High-ROI FinOps layer • Cuts monthly frontier API bills by ~58%.
● Seated
+ Add a tool that is not listed
We cannot check unlisted tools for license or speed. We use your answers for the cost and privacy checks, and list it by name in your export.
Socket 05Core Layer • Orchestrator
Multi-Agent Orchestration & State GraphiWhat Socket 05 DoesCoordinates multi-step agent workflows, tool handoffs, and typed state graphs.Use when chaining multiple agents/tools • Skip for single-agent coding.
● Seated
+ Add a tool that is not listed
We cannot check unlisted tools for license or speed. We use your answers for the cost and privacy checks, and list it by name in your export.
Socket 06Core Layer • Schema Out
Structured Output & Schema CompileriWhat Socket 06 DoesForces LLM outputs to strictly match your JSON/Pydantic types with zero parse crashes.Essential for production APIs & tool calling • Skip for plain chat outputs.
● Seated
+ Add a tool that is not listed
We cannot check unlisted tools for license or speed. We use your answers for the cost and privacy checks, and list it by name in your export.
3
Review and export your stack
Plain-English verdict, live architecture summary, and your 9-file production config bundle for the slots seated above.
Checking your stack...
Technical details RAM, GPU memory, license and budget checks behind the summary above
• Hardware Fit Verified: 2014 MB RAM & 16.0 GB VRAM fits inside your 24 GB GPU envelope.
Free Architectural Summary
Compiled ADR-2026-AI-STACK.md (Live Preview)
LIVE CANVAS
# ADR-2026-AI-STACK: Production AI Architecture Decision Record for `cline/cline` - **Status**: PROPOSED (Curated via PulseForge 12-Slot Deterministic BOM Engine) - **Team & Workload Profile**: `15` Engineers | `150,000` Monthly Agent Sessions | Mode: `HIPAA_AIRGAPPED` - **Configured Stack Slots**: `11` Active Layers | `1` Intentionally Omitted Layers (Lean YAGNI Setup) - **Total Monthly TCO**: `$519.00/mo` (vs. `$1,396.92/mo` Unoptimized Baseline -> **Saves `$877.92/mo`**) - **Concurrent Local RAM Footprint:** `2014 MB` | **Cumulative p50 Overhead:** `65.3 ms` | **Min Local VRAM:** `16.0 GB` ## 1. Curated Active Bill of Materials (BOM) & License Audit | Slot # | Architectural Layer | Selected Tool (`owner/repo`) | License | Cloud Egress | p50 Overhead | RAM (50 Agents) | Est. Monthly Cost | |---:|---|---|---|---|---:|---:|---:| | 1 | **AI Coding IDE & Editor** | `cline/cline` (Cline (BYOK IDE Agent)) | `Apache-2.0` | Zero-Egress Local/Self-Host | `9.5 ms` | `210 MB` | `$97.50/mo` | | 2 | **Autonomous Terminal & SWE Agent** | `Aider-AI/aider` (Aider (Tree-Sitter RepoMap)) | `Apache-2.0` | Zero-Egress Local/Self-Host | `7.8 ms` | `125 MB` | `$127.50/mo` | | 3 | **Agent Harness & Context OS** | `UKGovernmentBEIS/inspect_ai` (Inspect AI Safety & Eval Harness) | `MIT` | Zero-Egress Local/Self-Host | `9.2 ms` | `260 MB` | `$60.00/mo` | | 4 | **LLM Gateway, KV-Cache & FinOps Router** | `tensorzero/tensorzero` (TensorZero (Rust 0.8ms Gateway)) | `Apache-2.0` | Zero-Egress Local/Self-Host | `0.8 ms` | `64 MB` | `$48.00/mo` | | 5 | **Multi-Agent Orchestration & State Graph** | `pydantic/pydantic-ai` (PydanticAI Type-Safe Agents) | `MIT` | Zero-Egress Local/Self-Host | `2.4 ms` | `220 MB` | `$46.50/mo` | | 6 | **Structured Output & Schema Compiler** | `dottxt-ai/outlines` (Outlines FSM Constrained Decoding) | `Apache-2.0` | Zero-Egress Local/Self-Host | `1.5 ms` | `95 MB` | `$30.00/mo` | | 7 | **Agent Memory, Vector & Hybrid RAG Engine** | `lancedb/lancedb` (LanceDB (Embedded Disk-Native Rust)) | `Apache-2.0` | Zero-Egress Local/Self-Host | `2.1 ms` | `145 MB` | `$21.00/mo` | | 8 | **Code Sandbox, MicroVM & Browser Runtime** | `daytonaio/daytona` (Daytona Self-Hosted Agent Workspace) | `Apache-2.0` | Zero-Egress Local/Self-Host | `18.0 ms` | `310 MB` | `$54.00/mo` | | 9 | **MCP Server Runtime & Tool-Poisoning Firewall** | `invariantlabs-ai/mcp-scan` (Invariant MCP-Scan Firewall) | `Apache-2.0` | Zero-Egress Local/Self-Host | `1.4 ms` | `60 MB` | `$15.00/mo` | | 10 | **50-Probe OWASP Red-Team & Regression CI Gate** | `gitleaks/gitleaks` (Gitleaks Pre-Commit Secret Gate) | `MIT` | Zero-Egress Local/Self-Host | `0.6 ms` | `45 MB` | `$7.50/mo` | | 12 | **Local Inference & QLoRA / GRPO Fine-Tune Engine** | `ggml-org/llama.cpp` (llama.cpp (Zero-Egress GGUF Q4_K_M)) | `MIT` | Zero-Egress Local/Self-Host | `12.0 ms` | `480 MB` | `$12.00/mo` | ## 1B. Intentionally Omitted Layers (Lean YAGNI Justification) - **Slot 11 (Voice, WebRTC & Realtime Multimodal Agent)**: `OMITTED BY DESIGN` - Text/code-only workload; WebRTC voice & realtime audio pipeline not required.
🔒 Gated Pro Execution Suite ($9.99 / $49)
9-File Ready-to-Run Production Bundle
# docker-compose.ai-stack.yml - Generated for cline/cline (11 Active Slots | 1 Omitted | HIPAA_AIRGAPPED)
version: '3.9'
services:
layer_1_ide:
image: ghcr.io/cline/cline:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=AI Coding IDE & Editor
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 210M
layer_2_terminal_agent:
image: ghcr.io/aider-ai/aider:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Autonomous Terminal & SWE Agent
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 128M
layer_3_agent_harness:
image: ghcr.io/ukgovernmentbeis/inspect_ai:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Agent Harness & Context OS
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 260M
layer_4_gateway_router:
image: ghcr.io/tensorzero/tensorzero:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=LLM Gateway, KV-Cache & FinOps Router
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 128M
layer_5_orchestration:
image: ghcr.io/pydantic/pydantic-ai:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Multi-Agent Orchestration & State Graph
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 220M
layer_6_structured_output:
image: ghcr.io/dottxt-ai/outlines:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Structured Output & Schema Compiler
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 128M
layer_7_memory_vector:
image: ghcr.io/lancedb/lancedb:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Agent Memory, Vector & Hybrid RAG Engine
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 145M
layer_8_sandbox_browser:
image: ghcr.io/daytonaio/daytona:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Code Sandbox, MicroVM & Browser Runtime
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 310M
layer_9_mcp_firewall:
image: ghcr.io/invariantlabs-ai/mcp-scan:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=MCP Server Runtime & Tool-Poisoning Firewall
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 128M
layer_10_eval_redteam:
image: ghcr.io/gitleaks/gitleaks:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=50-Probe OWASP Red-Team & Regression CI Gate
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 128M
layer_12_local_finetune:
image: ghcr.io/ggml-org/llama.cpp:latest
restart: unless-stopped
environment:
- PULSEFORGE_SLOT=Local Inference & QLoRA / GRPO Fine-Tune Engine
- COMPLIANCE_MODE=hipaa_airgapped
deploy:
resources:
limits:
memory: 480M
docker-compose.ymlNon-conflicting container ports
litellm-cost-router.yaml65/35 FinOps -58% token router
CLAUDE.md + .cursorrulesAST context pruner & IDE rules
ai_eval_gate.yml + gitleaksOWASP LLM01 & MCP CI gate
UNSAVED ARCHITECTURE VAULT
11 Active • 1 Omitted
Export Your Curated Stack Bundle (Pay Once, Owned Forever)
Browse rankings and configure your 12 slots free. Unlock the compiled execution configs, router rules, and CI gates below:
- [FREE PREVIEW] 12-Slot Studio, +Add to Stack Bag & Live ADR Summary
- [GATED • $9.99 BOM .ZIP] Compiled docker-compose.yml, LiteLLM 65/35 Router, CLAUDE.md & CI Gates
- [GATED • $49 BUILDER KIT] Full 9-File .ZIP + Verified README Shield + Weekly Price/CVE Drift Alerts
[+] Maintainer / DevTool Founder: Test & Verify Your Repository in Any of the 12 BOM SlotsPreview Compatibility & Claim Slot Placement ▾
Preview how your tool pairs with upstream/downstream stack layers, or verify your repository's p50 latency, RAM, and license telemetry for default slot placement.