Prompts

Claude Console Guide for API Users

ClaudeAIHub. For official Anthropic Console documentation, visit platform.claude.com.

The Anthropic Console — accessed at platform.claude.com — is the web-based developer interface for the Claude API. It is the primary place to create API keys, test prompts before building them into code, monitor API usage, and manage your account settings. If you are building with the Claude API, the Console is where you start.

What the Console Is For

The Console is a developer tool, not the same as claude.ai. Where claude.ai is the conversational app for everyday users, the Console is for developers and organizations who want API access, programmatic usage, and account management.

Main areas of the Console include:

  • Workbench: A prompt testing interface where you can try requests against any current Claude model, adjust parameters (temperature, max tokens, system prompt), and inspect raw API responses — all without writing code.
  • API Keys: Create, name, and revoke API keys. Each key is shown only once at creation — after that it cannot be retrieved, only rotated.
    • Usage dashboard: Monitor token consumption across models, track spend over time, and set usage alerts.
  • Billing: Manage payment methods, view invoices, and review current billing tier.
  • Organization and workspace settings: Manage team members, permissions, and workspace-level settings if using an organization account.

Getting Started: Basic Developer Workflow

If you are new to the Claude API, this is the typical starting sequence in the Console:

  1. Sign in at platform.claude.com using your Anthropic account (same login as claude.ai).
  2. Go to Settings → API Keys and create a new API key. Copy it immediately — it will not be shown again.
  3. Store the key securely as an environment variable (see the Claude API Key Guide for storage best practices).
  4. Open the Workbench to draft and test your system prompt and message structure before writing code.
  5. Select a model in the Workbench — Opus 4.8 for complex tasks, Sonnet 4.6 for a balance of speed and intelligence, Haiku 4.5 for fast and cost-efficient requests.
  6. Adjust parameters: Set max tokens, temperature, and optionally a system prompt. Run test requests and review the output.
  7. Copy the request format into your application code using the Python or TypeScript SDK.
  8. Monitor usage in the usage dashboard to catch unexpected volume or cost spikes early.

Using the Workbench for Prompt Testing

The Workbench is one of the most useful parts of the Console for developers. It lets you test and refine prompts interactively before encoding them in your application:

  • System prompt editor: Write and test your system instructions in isolation before adding user messages.
  • Model selector: Switch between available Claude models to compare output quality and cost without changing code.
  • Parameter controls: Adjust max tokens and temperature to see how they affect output style and length.
  • Response inspection: View raw token counts (input, output, cached) in the response metadata — useful for cost estimation and prompt caching verification.
  • Export to code: Some views allow you to export the current Workbench configuration as Python or cURL code.

Prompt Improvement in the Console

The Console has included prompt improvement tooling that helps developers optimize their prompts. This tooling analyzes your prompt and suggests structural improvements — clearer instructions, better role framing, or more specific output format guidance. The goal is to reduce the number of test iterations needed to get reliable output from the model.

Some useful practices when using the Console for prompt development:

  • Start with a clear system prompt that defines Claude’s role, constraints, and output format.
  • Test edge cases in the Workbench before finalizing a prompt for production.
  • Use the response token count to estimate costs at scale — multiply output tokens per request by your expected request volume to estimate monthly API spend.
  • Keep a version log of your prompts outside the Console, as the Console does not maintain a version history of Workbench sessions.

Security Notes for Console Users

  • Protect your API keys. Never share API keys in public repositories, chat messages, or client-side code. If a key is compromised, revoke it immediately in the Console.
  • Monitor usage alerts. Set a usage alert in the Console billing settings to get notified if your spend exceeds a threshold — this can catch runaway API calls or unauthorized use early.
  • Do not test with production data. Avoid pasting sensitive customer data, confidential documents, or personal information into the Workbench for testing. Use anonymized or synthetic data instead.
  • Use workspace-level permissions. If your organization uses multiple workspaces, limit API key creation permissions to team members who need them.

Console vs claude.ai: Key Differences

FeatureAnthropic Console (platform.claude.com)Claude App (claude.ai)
AudienceDevelopers and API usersGeneral users
API key managementYesNo
Prompt testing (Workbench)YesNo (conversations only)
Usage and billing dashboardYesNo
Model parameter controlYes (temperature, max tokens, etc.)Limited
Conversation historyWorkbench sessions onlyFull conversation history
CostPay-per-token (API pricing)Free or subscription plan

Related Resources