# Kosuke Codex API Evaluation

**Category:** Coding Agent
**Score:** 74 (C)

This is the agent-readable Devtool Arena evaluation page for Kosuke on Codex API.

## Summary

| Metric | Value |
|--------|-------|
| Status | completed |
| Overall score | 74 |
| Grade | C |
| Eval score | 70 |
| Discovery score | 86 |
| Cost | $0.35 |
| Runtime | 2m 36s |
| Tool calls | 18 |
| Errors | 4 |
| Tokens used | 1549093 |
| Success | Yes |

## Agent-Readiness Checklist

| Signal | Evidence |
|--------|----------|
| Context7 | Not found |
| llms.txt | https://docs.kosuke.ai/llms.txt |
| MCP server | https://docs.kosuke.ai/api |
| Typed SDK | https://registry.npmjs.org/@kosuke-ai/cli |
| OpenAPI | https://app.kosuke.ai/api/openapi.json |
| Agent skills | https://docs.kosuke.ai/skills |
| CLI | https://docs.kosuke.ai/api |

## Evaluation Prompt

````text

# Task
Using https://docs.kosuke.ai/api, complete the following task:

Using the Kosuke API, build me a script that delegates a small coding task to a Kosuke coding agent and collects the agent's result:

1. Use the official Kosuke Python SDK if one exists; otherwise use the documented REST API with Python `requests`
2. If the API needs a project, repository, or workspace to run an agent in, list the ones the credentials can reach and use the first one returned
3. Start a new agent session, task, or run with exactly this prompt:
   "Write and run a Python script that prints the SHA-256 hex digest of the UTF-8 string `lightsage-coding-agent-eval`. Reply with only the 64-character hex digest."
4. Poll until the agent finishes, waiting at most 10 minutes, and exit with an error if it fails, stops to ask a question, or times out
5. Read the agent's final reply and extract the 64-character hex digest from it
6. Print the result as JSON with fields: session_id, status, digest

The digest must come from the agent's reply. Do not compute it locally or hard-code it. 

## Execution
After creating the script, run it to verify it works:
```bash
cd /home/daytona/app && python <your_script>.py
```

The script should print output to stdout. If there are errors, debug and fix them until it runs successfully.

````

## Grader Results

| Check | Passed | Weight | Score | Details |
|-------|--------|--------|-------|---------|
| Cost | Yes | 0 | 0 | $0.35 ($0.20–0.40, good) |
| Time | Yes | 0 | 0 | 156s (2–3min, good) |
| Efficiency | Yes | 0 | 0 | 18 tool calls (≤20, acceptable) |
| Found Docs | Yes | 0 | — | Touched docs at docs.kosuke.ai: yes (12 matching tool calls) |
| Zero Errors | No | 0 | — | Tool outputs with errors/tracebacks: 4 |
| Syntax Valid | Yes | 0 | — | Code compiles/parses correctly |
| Created Files | Yes | 0 | — | Generated files: 1 (need ≥1) |
| Full Execution | Yes | 0 | — | Method: full_execution, Exit: 0 |

## Run Artifacts

| Artifact | Value |
|----------|-------|
| Conversation turns | 22 |
| Tool call traces | 18 |
| Generated files | 1 |
| Exit code | 0 |
| Completed at | 2026-10-05T18:40:18.406215+00:00 |

### Generated Files

- /home/daytona/app/delegate_kosuke.py

## Related Pages

- [Codex API leaderboard](/leaderboard/codex/api)
- [Compare Codex API companies](/leaderboard/codex/api/compare)
- [Agent Landscape](/leaderboard/discoverability)

Canonical URL: https://devtoolarena.com/codex/api/kosuke