openclaw - 💡(How to fix) Fix [Bug]: OpenRouter provider uses wrong base URL, causing all model calls to fail silently [2 comments, 2 participants]

Bug type

Behavior bug (incorrect output/state without crash)

Beta release blocker

No

Summary

Version: 2026.4.15 (041266a) Platform: Linux 6.14.8-2-pve (x64) · Node.js 22.22.2 Provider: OpenRouter

Summary

On first gateway start, OpenClaw generates ~/.openclaw/agents/main/agent/models.json with an incorrect base URL for the openrouter and arcee providers:

"baseUrl": "https://openrouter.ai/v1"

The correct OpenRouter API base URL is:

"baseUrl": "https://openrouter.ai/api/v1"

https://openrouter.ai/v1 returns the OpenRouter website HTML (200 OK), not the API. The HTTP call succeeds from a networking standpoint, so no error is raised, but the response contains no model output.

Symptoms

Every message via any channel (Telegram, etc.) returns:

⚠️ Agent couldn't generate a response. Please try again.

Gateway logs show:

incomplete turn detected: stopReason=stop payloads=0 — surfacing error to user

Session files record the assistant message with empty content and zero tokens:

{
  "role": "assistant",
  "content": [],
  "usage": { "input": 0, "output": 0, "totalTokens": 0 },
  "stopReason": "stop"
}

This affects all models on the OpenRouter provider regardless of model tier (free or paid), thinking mode, or tools configuration. The issue persists across a full uninstall/reinstall because models.json is regenerated with the wrong URL on every fresh gateway start.

Root Cause

models.json is written by the gateway on first start. The generated file contains:

{
  "providers": {
    "openrouter": {
      "baseUrl": "https://openrouter.ai/v1",   // ← wrong
      ...
    },
    "arcee": {
      "baseUrl": "https://openrouter.ai/v1",   // ← wrong
      ...
    }
  }
}

A GET/POST to https://openrouter.ai/v1/chat/completions returns a Next.js HTML page. The OpenAI-compatible completions endpoint is at https://openrouter.ai/api/v1/chat/completions.

Fix

In the code that generates models.json, change the OpenRouter base URL from:

https://openrouter.ai/v1

to:

https://openrouter.ai/api/v1

Workaround

Manually edit ~/.openclaw/agents/main/agent/models.json and replace both occurrences of https://openrouter.ai/v1 with https://openrouter.ai/api/v1, then restart the gateway:

sed -i 's|https://openrouter.ai/v1|https://openrouter.ai/api/v1|g' \
  ~/.openclaw/agents/main/agent/models.json
systemctl --user restart openclaw-gateway

Note: This file is regenerated on gateway start, so the workaround may need to be reapplied after an OpenClaw upgrade.

Steps to reproduce

1. Fresh install: npm install -g openclaw
2. Configure OpenRouter provider with a valid API key
3. Start the gateway: systemctl --user start openclaw-gateway
4. Send any message via a configured channel
5. Observe: "⚠️ Agent couldn't generate a response" on every message
6. Check ~/.openclaw/agents/main/agent/models.json — baseUrl is https://openrouter.ai/v1
7. Verify: curl https://openrouter.ai/v1/chat/completions returns HTML, not JSON

Expected behavior

Model reply expected.

Actual behavior

"⚠️ Agent couldn't generate a response" on every message

OpenClaw version

026.4.15 (041266a)

Operating system

Ubuntu 24 Node.js 22.22.2

Install method

npm

Model

openrouter/free

Provider / routing chain

openrouter

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response