estimate

/estimate — Pre-build capacity & cost estimator

You are generating an interactive HTML capacity-planning report for the user's proposed architecture(s), before any code is written.

Paths

This command ships as a Claude Code plugin. Two distinct path bases are used:

• ${CLAUDE_PLUGIN_ROOT} — set by Claude Code to this plugin's installed location. All plugin-bundled assets (catalog/*.json, template/base.html, docs/, examples/) live here. Always read assets via this prefix, e.g. ${CLAUDE_PLUGIN_ROOT}/catalog/aws.json, ${CLAUDE_PLUGIN_ROOT}/template/base.html.
• The user's current working directory (no prefix) — output goes here, in ./.claude/estimates/<timestamp>.html. This keeps generated reports next to the project the user is sizing, not buried in the plugin's install dir.
• The user's project files you scan (terraform/k8s/migrations/etc.) are also in the user's cwd, with no prefix.

When in doubt: read plugin-relative; write cwd-relative.

Inputs

The user passes a free-form description as $ARGUMENTS. Examples:

• REST API for user profile + recent orders, 5k RPS, AWS
• Social feed, 2M DAU, AWS vs GCP
• Payments, 500 TPS peak, AWS
• WebSocket chat, 100k connections, compare AWS vs GCP
• Event ingestion 500 MB/s, Kafka vs Kinesis

Bare invocation — /estimate with no arguments

If $ARGUMENTS is empty or whitespace-only, do not immediately ask the generic missing-load follow-up. Instead, autonomously scan the project for infra/workload signals first, synthesize a one-line hypothesis, and ask one focused confirm-or-correct question.

Scan list (in order, stop early if signal is strong):

package.json, pnpm-lock.yaml, yarn.lock         → deps reveal workload type:
                                                    socket.io / ws / @nestjs/websockets → WS
                                                    kafkajs / @nestjs/microservices / pulsar → Stream
                                                    express / fastify / nestjs / koa / next → REST (default)
requirements.txt, pyproject.toml, Pipfile       → flask/django/fastapi/aiohttp → REST
                                                    aiokafka / faust → Stream
                                                    websockets / channels → WS
go.mod                                          → gin/echo/fiber/chi → REST; nhooyr/gorilla → WS; segmentio/sarama → Stream
Gemfile, composer.json, build.gradle, pom.xml   → framework signals same as above
Dockerfile, docker-compose*.yml                 → exposed ports, base images
serverless.yml, template.yaml (SAM), samconfig  → existing Lambda layout, runtime, memory
terraform/**/*.tf, *.tf                         → aws_instance / aws_db_instance / aws_lb_*,
                                                    google_compute_instance, azurerm_*,
                                                    aws_lambda_function, aws_dynamodb_table
k8s/**/*.yaml, kustomize/**/*.yaml, helm/       → Deployment containers, requests/limits, HPA min/max
prisma/schema.prisma, drizzle/, sqlx/migrations,
migrations/*.sql, schema.sql, models.py         → DB engine, table shape (users/posts → social;
                                                    products/orders → ecom; messages/rooms → chat)
.env.example, config/*.yml                      → REDIS_URL / DATABASE_URL / KAFKA_BROKERS hints
README.md, CLAUDE.md (top of repo)              → domain hint