design_automation_stack
Route a vague workflow into a concrete automation architecture.
Free public beta · Remote MCP + API for practical AI automation design
Turn “automate this” into a real agent workflow.
Most agent requests fail because they skip tool routing, inputs, permissioned account routes, human boundaries, and QA. This MCP returns the stack, command, permissioned account route, dry-run path, validation checks, and recovery plan before an agent builds another empty shell.
Try the architectFeedback issueStar on GitHubLaunch kit
https://ai-automation-operating-pack.vercel.app/api/mcpInputPlain-language workflow
“I need Gmail + Docs follow-up automation” or “Build an Obsidian learning cleanup flow.”
OutputTool stack + prompt
Installed plugin/MCP/skill route, permissioned connector plan, copy-paste command, dry-run, QA, and boundary.
StandardNo-shell validation
PASS only when the workflow reaches a usable sample result with evidence.
StatusFree public beta
No payment flow. Feedback, issues, stars, and real workflow examples decide the next version.
Live API preview
Describe one workflow.
The response is deterministic and free: no external model call, no secret, no unpermissioned account action.
MCP tools
generate_no_shell_prompt
Create a copy-paste command for Codex, Claude, Gemini, or another agent.
recommend_agent_tools
Choose plugins, MCPs, skills, permissioned account routes, and live-action boundaries for the job.
audit_automation_plan
Score an existing plan for shell-risk and missing execution pieces.
build_validation_pack
Return dry-run, QA, audit, and PASS criteria.
build_customer_intake
Generate a 10-minute intake for non-developers.
build_m2m_package_contract
Return MCP/API surfaces, delivery artifacts, buyer handoff, and paid-readiness gates for package integrators.