Disclosure and availability: DaoXE is a multi-model multi-protocol API gateway we operate (Chat Completions, Responses, Anthropic Messages, and related surfaces). It is listed on models.dev as
daoxewithapi=https://daoxe.com/v1. It is not available in mainland China. This article is for developers in regions allowed by the service terms.
You paste a custom base URL into a coding agent.
It fails.
The failure is usually not “the gateway is down.” It is usually a pre-flight problem: protocol, path, key name, or model id.
This is the 10-minute checklist I run before I open the agent UI.
Series context
- Smoke test
- Multiprotocol
- Client setup
- curl OK, IDE fails
- Claude protocol
- models.dev + /models
- This post: pre-flight for coding agents
Minute 1–2: Name the client dialect
| Client style | Usually needs |
|---|---|
| OpenAI-compatible coding tools | Chat Completions + Bearer key + base with /v1
|
| Anthropic-shaped tools | Messages + x-api-key + anthropic-version
|
| Catalog-driven tools | provider key (e.g. daoxe on models.dev) + env name |
Do not configure Claude Messages settings for a Chat Completions-only client.
Minute 3–4: Auth + host with curl
export DAOXE_API_KEY="your_api_key"
export DAOXE_BASE_URL="https://daoxe.com/v1"
curl --fail-with-body --show-error --silent \
-H "Authorization: Bearer $DAOXE_API_KEY" \
"$DAOXE_BASE_URL/models"
Pass: HTTP 200 and model ids.
Fail: fix key / host before opening the agent.
Minute 5–6: One generation, exact model id
export DAOXE_MODEL="paste_exact_id_from_models"
curl --fail-with-body --show-error --silent \
-H "Authorization: Bearer $DAOXE_API_KEY" \
-H "Content-Type: application/json" \
-d "{
\"model\": \"$DAOXE_MODEL\",
\"max_tokens\": 8,
\"messages\": [{\"role\":\"user\",\"content\":\"Reply with OK.\"}]
}" \
"$DAOXE_BASE_URL/chat/completions"
Only a green curl here is permission to touch the agent config.
Minute 7–8: Clone the triple once
| Field | Example |
|---|---|
| Base URL | https://daoxe.com/v1 |
| Key | env / secret store, not committed |
| Model | exact id from /v1/models now |
Public notes for common clients:
Minute 9–10: One agent only
Do not fan out to five tools.
Configure one agent. Confirm a one-line reply. Then clone the working triple.
Common fails after this checklist
| Fail | Usually means |
|---|---|
Double /v1
|
client already appends path |
| model_not_found | stale id from a blog |
| 401 in agent, 200 in curl | wrong key field / env name |
| Claude client fails after chat works | wrong protocol surface |
Soft CTA
Product side of this writeup:
https://daoxe.com/?utm_source=devto&utm_medium=organic&utm_campaign=global_launch_agent_checklist
Comment with client name + which step failed (1–4). Do not paste keys.
Top comments (0)