specification: FinOps Framework specificationVersion: '1.0' alignedWith: framework: FinOps Foundation Framework frameworkUrl: https://www.finops.org/framework/ dataSpec: FOCUS dataSpecVersion: '1.3' dataSpecUrl: https://focus.finops.org/focus-specification/v1-3/ provider: Plandex providerId: plandex publisherName: PlandexAI Inc. serviceCategory: Developer Tools / AI Coding Agent created: '2026-05-29' modified: '2026-05-29' tags: - AI Coding Agent - Developer Tools - FinOps - Cost Management - FOCUS - LLM - Open Source description: >- FinOps framework definition for Plandex. Plandex itself is MIT-licensed and free; the hosted Plandex Cloud (now winding down as of 2025-10-03) charged a subscription plus pass-through-with-markup model credits in Integrated Models Mode, and a flat subscription in BYO API Key Mode. The dominant ongoing cost for users running Plandex (cloud or self-hosted) is upstream model-provider consumption (OpenAI, Anthropic, OpenRouter, Google, Azure, Bedrock, DeepSeek, Perplexity, Relace). This document maps that surface to the FinOps Foundation framework and FOCUS data spec so consuming organizations can build cost visibility, allocation, optimization, and accountability around Plandex. status: historical-and-active principles: - name: Visibility description: >- Surface per-plan, per-branch, per-user, and per-model spend through the `plandex usage` command (and historical Plandex Cloud billing settings page) so engineering and finance can see where AI coding spend is going in near real time. - name: Allocation description: >- Tag every model call with the consuming user, plan, branch, model pack, role (planner / coder / builder / summarizer / verifier / context-loader), and upstream provider so cost can be charged back to teams. - name: Optimization description: >- Use the daily, cheap, oss, and planner-specialized model packs to trade capability for cost. Use smart-context, auto-load-context, and the tree-sitter file map to avoid sending unnecessary tokens. Use the built-in prompt caching across OpenAI, Anthropic, and Google models to reduce input-token cost on multi-step plans. - name: Accountability description: >- Establish a Plandex budget owner per org or team. Use auto-recharge, monthly budgets, and email notification thresholds (or the equivalent on self-hosted deployments) to keep spend predictable. domains: - name: Understand Usage and Cost capabilities: - Data Ingestion - Allocation - Reporting and Analytics - Anomaly Management - name: Quantify Business Value capabilities: - Planning and Estimating - Forecasting - Budgeting - Benchmarking - Unit Economics - name: Optimize Usage and Cost capabilities: - Workload Optimization - Rate Optimization - Architecting for Cloud - Cloud Sustainability - Licensing and SaaS - name: Manage the FinOps Practice capabilities: - FinOps Practice Operations - FinOps Education and Enablement - Invoicing and Chargeback - Onboarding Workloads - Intersecting Disciplines billingModel: pricingCategory: Hybrid billingFrequency: Monthly billingCurrency: USD chargeCategories: - Usage - Purchase - Tax - Credit - Adjustment chargeFrequency: Recurring components: - name: Self-Hosted License description: MIT license — no Plandex-side fee. Upstream model-provider costs are billed by each provider directly. pricing: free - name: Plandex Cloud Subscription (BYO) description: $30/month flat subscription. Historical — Plandex Cloud winding down 2025-10-03. pricing: subscription - name: Plandex Cloud Subscription (Integrated) description: $45/month subscription including $20 of non-expiring credits. Historical — Plandex Cloud winding down 2025-10-03. pricing: subscription-plus-credits - name: Plandex Credits description: Pay-as-you-go usable against integrated model providers; charged at provider price plus a small markup. Historical. pricing: pass-through-with-markup - name: Upstream Provider Charges description: Direct charges from OpenAI, Anthropic, OpenRouter, Google AI Studio, Vertex AI, Azure OpenAI, AWS Bedrock, DeepSeek, Perplexity, Relace based on token usage. pricing: per-token / per-request (provider-defined) focusColumns: ServiceName: Plandex ServiceCategory: Developer Tools / AI Coding Agent ProviderName: PlandexAI Inc. PublisherName: PlandexAI Inc. InvoiceIssuerName: PlandexAI Inc. PricingCategory: Hybrid PricingUnit: subscription-month / credit-usd / provider-token BillingCurrency: USD ChargeCategory: Usage meters: - name: subscription_month description: Active months of Plandex Cloud subscription (BYO or Integrated). unit: month aggregation: sum dimensions: - mode - org - user - name: plandex_credit_usage description: Plandex credits spent against integrated model providers. unit: credit-usd aggregation: sum dimensions: - org - user - plan - branch - upstream_provider - upstream_model - role - name: plandex_credit_purchase description: Plandex credits purchased (one-off top-up or auto-recharge). unit: credit-usd aggregation: sum dimensions: - org - user - name: upstream_input_tokens description: Input tokens consumed against upstream model providers when running self-hosted (or BYO) Plandex. unit: token aggregation: sum dimensions: - upstream_provider - upstream_model - plan - branch - role - cached - name: upstream_output_tokens description: Output tokens generated by upstream model providers. unit: token aggregation: sum dimensions: - upstream_provider - upstream_model - plan - branch - role - name: model_response_count description: Count of model responses across plans (used by Cloud BYO trial cap). unit: response aggregation: sum dimensions: - plan - branch - role allocationTags: - org - user - plan - branch - model_pack - role - upstream_provider - environment optimizationLevers: - Use the `cheap` or `oss` model pack for routine tasks; reserve `strong` / `opus-planner` / `o3-planner` for high-value planning. - Enable `smart-context` and `auto-load-context` so only the necessary files for each step are sent to the model. - Use tree-sitter file maps instead of full file bodies for very large projects. - Take advantage of built-in context caching across OpenAI, Anthropic, and Google models to reduce input-token cost on multi-step plans. - Set monthly budgets and email notification thresholds (Plandex Cloud) or external provider budgets (self-hosted) to fail safe. - Use OpenRouter as the failover provider to absorb transient direct-provider errors without halting plans. maintainers: - FN: Kin Lane email: kin@apievangelist.com