These are free mock calculators. The score comes from the answers you enter — an illustrative estimate, not real metrics. Sign up to get real metrics from your real, connected data.
Get real metricsChoose the right AI models and plan how to ship.
Get model recommendations, reasoning settings, a cost estimate, and a development workflow for your project — computed deterministically from a sourced catalog, free and in your browser.
What this planner does
You describe a project, pick a workload, priorities, budget, and expected volume, and get a complete plan: which model to use for planning, implementation, verification, and documentation (with the right reasoning effort for each), a safe orchestration workflow, a low/base/high cost estimate, and an eight-stage delivery lifecycle. It helps individual builders pick models and avoid wasted retries, and helps engineering and AI-platform teams plan model usage, budgets, permissions, reviews, and delivery controls.
The planner
Context only — the plan is driven by your explicit selections below, not by parsing this text. 0/1500
Deterministic — no API key, nothing leaves your browser.
How the recommendations are made
A maintained catalog records each model’s exact API id, lifecycle, context limits, provider-specific reasoning parameters, and pricing — verified against official provider documentation with a source URL and a verification date. The engine applies your hard constraints first (allowed providers, data sensitivity), then ranks eligible models per role and priority with transparent, versioned heuristics. Prices and reasoning settings are provider-specific and never fabricated; unknown prices produce an incomplete estimate, never a $0. This is a starting recommendation, not a benchmark of your specific workload.
Frequently asked questions
- Who is the AI Orchestration Planner for?
- Two audiences. Individual builders and 'vibe coders' who want to pick models, set reasoning effort, avoid wasted retries, and ship safely. And enterprise engineering / AI-platform teams planning model usage, budgets, permissions, reviews, and delivery controls across a team.
- What does the free plan include?
- Everything: model and reasoning-effort recommendations for planning, implementation, verification, and documentation; an orchestration workflow with tools, permissions, limits, and human-approval boundaries; a low/base/high cost estimate; an eight-stage delivery lifecycle; and downloads (copy, Markdown, versioned JSON, and a starter agent-instruction document). No signup, no API key, and no paid model call are required to generate or export a plan.
- How are the model recommendations made?
- Deterministically. A maintained catalog records each model's exact API id, lifecycle, context limits, provider-specific reasoning parameters, and pricing — verified against official provider docs with a source URL and date. The engine applies your hard constraints first (allowed providers, sensitivity), then ranks the eligible models per role and priority with transparent, versioned heuristics. There is no AI call and no fabricated quality score.
- Why do the cost estimates vary?
- The estimate is a low/base/high range built from assumed tokens per step, your complexity and retry assumptions, and each model's published rates — reasoning/thinking tokens are billed at the output rate, and prompt-cache discounts are deliberately not applied. It is an estimate, not a quote or an enforced cap. If a model's price can't be verified the estimate is marked incomplete rather than shown as $0.
- Does the planner enforce anything — budgets, compliance, or the workflow?
- No. It PLANS. It does not execute agents, connect to your repository, enforce a runtime budget, or certify compliance. The controls it recommends (permissions, approval gates, verification) are proposals for your team or agent runtime to implement — the plan clearly separates 'proposed' from anything ShipReadyMetrics actually enforces.
- Reasoning effort: is 'high' the same across providers?
- No — reasoning settings are provider-specific. Anthropic exposes output_config.effort (low–max), OpenAI exposes reasoning_effort, and Google exposes a thinking level. The planner only ever suggests a value a given model actually supports, and never sends one provider's setting to another. A model with no configurable reasoning is shown as 'Not configurable'.
- Is a paid coding subscription the same as API credits?
- No. A coding-tool or IDE subscription (a seat fee) is different from API consumption. Selecting a paid subscription does not give you API credits or let the planner route that tool's internal model calls — the cost estimate is about model API spend only.