TOKENS

Paste text or drop a file. Compare tokens and costs across models.

302 tokens153 words1,056 chars

Code detected

302 tokens153 words1,056 chars

You are an AI agent that helps developers build applications. Your task is to:

1. Analyze the user's codebase structure
2. Identify areas where AI can add value
3. Suggest concrete implementations using TypeScript
4. Generate working code with proper error handling

When building an AI agent, consider these key principles:
- Keep the agent loop simple and predictable
- Use structured tool definitions with clear schemas
- Implement proper retry logic and timeout handling
- Log every step for debugging and observability
- Test with deterministic inputs before going live

Here is an example tool definition:

const tools = [
  {
    name: "search_codebase",
    description: "Search for files and code patterns in the repository",
    parameters: {
      type: "object",
      properties: {
        query: { type: "string", description: "The search query" },
        fileType: { type: "string", description: "Filter by file extension" }
      },
      required: ["query"]
    }
  }
];

Please analyze the attached codebase and provide recommendations.

Context window usage (sorted by size)

Gemini 2.5

0.03%

Sonnet 4.6

200K

0.15%

GPT-5.3

200K

0.15%

Opus 4.6

200K

0.15%

Llama 3.3

128K

0.24%

GPT-4o

128K

0.24%

Cost comparison (sorted by cheapest input)

Model	Input /1M	Output /1M	This text (in)	This text (out)
Llama 3.3	$0.2	$0.2	< $0.01	< $0.01
Gemini 2.5	$1.25	$5	< $0.01	< $0.01
GPT-4o	$2.5	$10	< $0.01	< $0.01
Sonnet 4.6	$3	$15	< $0.01	< $0.01
GPT-5.3	$4.81	$4.81	< $0.01	< $0.01
Opus 4.6	$15	$75	< $0.01	$0.02

Get Smarter About AI Dev

New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.

One email per weekReal code, not theoryFree forever