# Naive Claude Code MCP Evaluation

**Category:** Unified API
**Score:** 33 (D)

This is the agent-readable Devtool Arena evaluation page for Naive on Claude Code MCP.

## Summary

| Metric | Value |
|--------|-------|
| Status | completed |
| Overall score | 33 |
| Grade | D |
| Eval score | 33 |
| Discovery score | 35 |
| Cost | $0.16 |
| Runtime | 7m 41s |
| Tool calls | 9 |
| Errors | 1 |
| Tokens used | 241326 |
| Success | No |

## Agent-Readiness Checklist

| Signal | Evidence |
|--------|----------|
| Context7 | Not found |
| llms.txt | Not found |
| MCP server | Not found |
| Typed SDK | Not found |
| OpenAPI | Not found |
| Agent skills | Not found |
| CLI | Not found |

## Evaluation Prompt

````text

# Task
Using the Naive MCP server, run a web research workflow:

1. Use the available Naive MCP tools to search for "AI agent infrastructure trends 2026"
2. Return the top 3 results with title, URL, and snippet or equivalent summary fields
3. If the MCP tools support URL extraction, read the first returned URL and print a short extracted summary
4. Print all output clearly labeled for each step

Use ONLY Naive MCP tools for Naive operations. Do NOT use curl, raw HTTP, CLI commands, or SDK calls as substitutes.

You MUST use the Naive MCP server to accomplish this task.
Use the MCP tools for the actual product operations. Do NOT use curl, fetch, raw HTTP requests, CLI commands, or SDK calls as substitutes.

Use the MCP tools' help/descriptions to understand what each tool does.
If you need additional context, check the documentation: https://usenaive.ai/docs/mcp/overview

## Setup
Do NOT assume the Naive MCP server is already installed or already running in this sandbox.
Review the installation and authentication details below before starting the task.

### Installation
Transport: `sse`
This MCP server uses `sse` transport.
Claude Code should connect to: `https://api.usenaive.ai/mcp/sse`
Do not assume a local MCP package is already installed; the task depends on the remote server being reachable.
Bootstrap notes: The stored docs_url (https://usenaive.ai/docs/mcp/overview) returned HTTP 404 at verification time; no dedicated 'MCP overview' page currently exists on the site. Directly probing the stored server URL confirmed it is real and live: GET https://api.usenaive.ai/mcp/sse returns HTTP 401 with body {"error":{"code":"unauthorized","message":"Valid API key (nv_sk_…) or session token (nv_sess_…) required for MCP"}} and Naive-specific headers (x-naive-build, x-naive-cli-la*****/min), confirming this is a genuine, auth-gated Naive endpoint (not a 404/placeholder) and that GET is meaningful (consistent with SSE, unlike the sibling Vetta JSON-RPC endpoint at api.vetta.sh/v1/mcp which only accepts POST and 405s on GET). Naive is built on the 'Vetta' infrastructure (per https://usenaive.ai/docs/naive-and-vetta) and Vetta's documented API auth scheme is 'Authorization: Bearer <key>' with keys prefixed sk_live_/sk_ (env var VETTA_API_KEY). By the same convention, and per the stored hint, the Naive key is expected to be supplied as 'Authorization: Bearer <NAIVE_API_KEY>' (key format nv_sk_... per the live 401 body) -- this exact header name was not independently confirmed via a documented example for the naive-branded endpoint (only inferred by sending it and by cross-product convention), so treat the header name as verified-by-convention rather than doc-confirmed.

### Authentication
The following environment variables are already set in this environment: `NAIVE_API_KEY`
The MCP server should read them automatically when it starts.

## Execution
1. The harness has already written the launch configuration for `naive`. Start with the task directly; do not rewrite agent settings, `.mcp-runtime.json`, or build a separate MCP client unless the connected server still fails after a normal retry.
2. You do not need to enumerate every MCP tool up front; inspect only the relevant tools you need. Agent-facing MCP tool names may be namespaced like `mcp__naive__<tool_name>`.
3. Complete the task described above using MCP tools for the actual product operations.
4. If the server fails to start or authenticate, debug the documented setup and retry. Do not fall back to raw HTTP, SDK, CLI, or manual JSON-RPC/MCP calls as a substitute for exposed MCP tools.

Work in the /home/daytona/app directory.
If you use Bash during setup or debugging, limit it to installation/configuration work and then return to MCP tools for the task itself.
If there are errors, debug and fix them until the task runs successfully.

## Summary
When you are done, write a JSON summary file to /home/daytona/app/results.json with this structure:
{
  "operations": [
    {"step": 1, "description": "What you did", "mcp_tool": "tool_name", "success": true/false, "output_summary": "brief result"},
    ...
  ],
  "overall_success": true/false,
  "notes": ["any relevant notes about the execution"]
}

````

## Grader Results

| Check | Passed | Weight | Score | Details |
|-------|--------|--------|-------|---------|
| Task | No | — | — | — |
| Llms Txt | No | — | 0 | — |
| Tool Count | No | — | 0 | — |
| Auth Method | Yes | — | 15 | — |
| Official Mcp | Yes | — | 5 | — |
| Output Schema | No | — | 0 | — |
| Tool Stability | Yes | — | 5 | — |
| Tool Description Length | Yes | — | 10 | — |

## Run Artifacts

| Artifact | Value |
|----------|-------|
| Conversation turns | 11 |
| Tool call traces | 9 |
| Generated files | 1 |
| Exit code | — |
| Completed at | 2026-09-27T14:28:12.193269+00:00 |

### Generated Files

- /home/daytona/app/results.json

## Related Pages

- [Claude Code MCP leaderboard](/leaderboard/claudecode/mcp)
- [Compare Claude Code MCP companies](/leaderboard/claudecode/mcp/compare)
- [Agent Landscape](/leaderboard/discoverability)

Canonical URL: https://devtoolarena.com/claudecode/mcp/naive