References: Shipping a Claude application

Source material

Source curriculum (structural mirror, cited as further study):
• Anthropic Academy (https://anthropic.skilljar.com/):
    "Building with the Claude API" course (production sections)
  License: Anthropic Academy course content is account-gated;
    Clawdemy structurally mirrors the Academy's lesson progression
    as inspiration and cites it as further study. Every substantive
    claim in this lesson is verifiable against the public Anthropic
    documentation.

Primary public-doc anchors (every substantive claim verified
against):
• Anthropic, "Usage and Cost API" (the verbatim positioning,
   the two endpoints, Admin API key requirement, bucket widths,
   filter and group_by dimensions, 5-minute data-freshness caveat,
   Priority-Tier-not-in-cost-endpoint caveat, Claude Platform on
   AWS limitation, partner integrations)
  https://platform.claude.com/docs/en/manage-claude/usage-cost-api
• Anthropic, "Admin API" (the broader Admin API context;
   Admin key auth; organization vs individual accounts)
  https://platform.claude.com/docs/en/manage-claude/admin-api
• Anthropic, "Model deprecations" (the four lifecycle states
   verbatim, the at-least-60-days notice, the Console Export
   audit path, the temperature/top_p/top_k deprecation on
   Opus 4.7+, currently-deprecated and recently-retired tables)
  https://platform.claude.com/docs/en/about-claude/model-deprecations
• Anthropic, "Pricing" (per-model rates; Batch API discount;
   prompt-caching multipliers)
  https://platform.claude.com/docs/en/about-claude/pricing
• Anthropic, "Service tiers" (Priority Tier framing; how it
   interacts with the Cost endpoint)
  https://platform.claude.com/docs/en/api/service-tiers
• Anthropic, "Rate limits" (organization-level limits; how
   the Usage API helps you stay within them)
  https://platform.claude.com/docs/en/api/rate-limits
• Anthropic, "Migration guide" (per-model migration steps)
  https://platform.claude.com/docs/en/about-claude/models/migration-guide
• Anthropic, "Claude Code Analytics API" (the separate
   per-user breakdown for Claude Code workloads)
  https://platform.claude.com/docs/en/manage-claude/claude-code-analytics-api

Verbatim claims sourced from the public docs:
• Usage and Cost API positioning: "The Usage & Cost Admin API
   provides programmatic and granular access to historical API
   usage and cost data for your organization"
• Data freshness: "Usage and cost data typically appears within
   5 minutes of API request completion, though delays may
   occasionally be longer"
• Polling cadence: "The API supports polling once per minute
   for sustained use"
• Priority Tier caveat: "Priority Tier costs use a different
   billing model and are not included in the cost endpoint.
   Track Priority Tier usage through the usage endpoint instead"
• AWS limitation: "The programmatic Usage and Cost API endpoints
   are not currently available" on Claude Platform on AWS
• Lifecycle states (model-deprecations docs): "Active: The model
   is fully supported and recommended for use. Legacy: The model
   will no longer receive updates and may be deprecated in the
   future. Deprecated: The model is still functional but no
   longer recommended. Anthropic provides a recommended
   replacement and assigns a retirement date. Retired: The model
   is no longer available for use. Requests to retired models
   will fail"
• Notice window: "Anthropic notifies customers with active
   deployments for models with upcoming retirements, providing
   at least 60 days notice before model retirement for publicly
   released models"
• Parameter deprecation: temperature / top_p / top_k "Returns a
   400 error when set to a non-default value on Claude Opus 4.7
   and later, including Claude Opus 4.8"

Required attribution: "Based on the structure of the Anthropic
  Academy 'Building with the Claude API' course
  (https://anthropic.skilljar.com/). This lesson is an
  independent structural mirror in original prose; every
  substantive claim is verified against the public Anthropic
  Claude documentation at https://platform.claude.com/docs/.
  Anthropic does not endorse it."

Read this next

• Anthropic, “Usage and Cost API”. The canonical reference: every parameter, every group-by dimension, every filter, the full response schema, partner integrations, pagination.
• Anthropic, “Model deprecations”. The full lifecycle policy, the currently-deprecated and recently-retired tables, the API parameter deprecation table.
• Anthropic, “Migration guide”. Per-model migration steps. Read this when a deprecation notice lands or when a new flagship arrives.
• Anthropic, “Pricing”. The canonical per-model rate sheet. Cross-reference with cache-write and cache-read multipliers from L7.

Going deeper

A short, durable list. Each link is a deeper destination beyond Track 22.

• Track 21 (LLM Ops and Production), lesson 7 “LLMOps.” The provider-agnostic playbook for the discipline as a whole: observability, evaluation in production, prompt versioning, cost and latency monitoring, regression testing. T22 L12 here is the Anthropic-specific instance of those five operational pillars.
• Track 20 (AI Agents and Tool Use). The full track-level treatment of agent design, harness engineering, and the operational discipline of running agents in production over time. Pairs well with T22 L8-L11.
• Anthropic Claude Cookbook. Worked notebook recipes for many of the patterns this lesson references (cache-efficiency dashboards, cost-attribution reports, budget alerts, CSV exports).
• Anthropic engineering, “Building Effective AI Agents”. The L9 patterns essay; relevant for the per-pattern cost and rollout disciplines this lesson tied off.

Where this connects inside the track

The closer that ties L1-L11 together:

• Lesson 1 (first call): the iterate-the-content-array discipline + the usage object; per-call logging shape this lesson formalizes starts from L1.
• Lesson 2 (production patterns): request_id logging, stop_reason dispatch, streaming, batches API. All load-bearing for the rollout + cost-monitoring disciplines here.
• Lesson 3 (model selection): the four cost levers + the date-pinning discipline + the deprecation-policy-aware production posture all start here.
• Lesson 4-6 (tools layers): the tool inventory is the safety surface area in the rollout checklist; denylist destructive ops; sandbox computer-use; MCP servers behind feature flags.
• Lesson 7 (caching + context management): cache-hit ratio is the headline cost metric this lesson tracks; compaction billing surfaces in usage.iterations; the cache-the-system-prompt-end pattern keeps compaction affordable.
• Lesson 8 (agent loop): hard max_iterations cap + explicit stop_reason dispatch + tool-inventory-as-surface-area are all in the rollout checklist.
• Lesson 9 (six patterns): pattern by decision tree (simplest that fits) is in the checklist; pattern 5 (evaluator-optimizer) is the loop for the eval-set discipline.
• Lesson 10 (Skills + Claude Code): durable Skills + CLAUDE.md in the checklist; security audit on third-party Skills is the per-rollout discipline.
• Lesson 11 (Subagents + Managed Agents): the right-primitive-per-workload decision is in the checklist; ZDR posture per primitive is the data-retention check; per-subagent model is a cost lever.
• And out to T21 L7 (“LLMOps”): the deeper provider-agnostic playbook for the five disciplines here.

The track is closed at L12. From here the natural continuations are T21 (operational discipline for the application as a whole) and T20 (deeper agent design).