Reference: Verify Your Connection

Before configuring an IDE or agent, confirm your key and endpoint work with a raw curl. This isolates connectivity/auth problems from agent-config problems.

On Windows native, run these from PowerShell (curl.exe is bundled with Windows 10+). On WSL/macOS/Linux, use a normal terminal.

1. Identity check (/whoami)

curl -s https://router.neosmith.ai/whoami \
  -H "Authorization: Bearer sk-plus-yourname-xxxxxx"

Expected: a JSON body with your dev slug, org, tier, and 30-day cap usage. A 401 means the key is wrong or missing.

2. Model list (/v1/models)

curl -s https://router.neosmith.ai/v1/models \
  -H "Authorization: Bearer sk-plus-yourname-xxxxxx"

Expected: a list including neosmith.intelligent-pro, neosmith.intelligent-basic, neosmith.intelligent-lite.

3. OpenAI-format chat (/v1/chat/completions)

curl -s https://router.neosmith.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-plus-yourname-xxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "neosmith.intelligent-pro",
    "messages": [{"role": "user", "content": "Say hello in one word."}],
    "max_tokens": 20
  }'

Expected: a normal OpenAI chat completion, plus a neosmith_meta field.

4. OpenAI Responses format (/v1/responses) — for Codex

curl -s https://router.neosmith.ai/v1/responses \
  -H "Authorization: Bearer sk-plus-yourname-xxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "neosmith.intelligent-pro",
    "input": "Say hello in one word.",
    "max_output_tokens": 20
  }'

Expected: a Responses-API body with an output array.

5. Anthropic-format message (/v1/messages) — for Claude Code

curl -s https://router.neosmith.ai/v1/messages \
  -H "x-api-key: sk-plus-yourname-xxxxxx" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4",
    "max_tokens": 20,
    "messages": [{"role": "user", "content": "Say hello in one word."}]
  }'

Expected: an Anthropic-format message with content blocks.

---

Interpreting failures

Result	Meaning	Fix
401 Unauthorized	Bad/missing key	Check the key; ensure no trailing whitespace
400 Unknown model	Model name not accepted	Use a neosmith.* SKU or claude-* id, not gpt-4o
Connection refused / timeout	Network/firewall	Ensure outbound HTTPS to router.neosmith.ai:443 is allowed
404 on /v1/chat/completions	Wrong base URL	Include /v1 for OpenAI-format clients
Works in curl, fails in IDE	Agent config issue	Recheck the agent’s base URL + key; see the per-agent guide