#auth-setup
Sorted by views, then solution_desc, solution, and root_cause length (desc).
2548 issues
When running CLI commands like `openclaw browser status`, the connection to the local gateway consistently fails with: `Error: gateway closed (1000 normal closure): no close reason` Interestingly, the command works perfectly when the `--json` flag is added. This suggests a race condition or event-loop blockage caused by the terminal progress indica tor (spinner).
The `openai-codex` provider (targeting `chatgpt.com/backend-api/codex/responses`) does not cap the `instructions` field before sending. When workspace bootstrap files + system prompt + tool schemas exceed ~32 KiB, the API returns `{"detail":"Bad Request"}` (400). This makes GPT-5.4 unusable for agents with larger workspaces.
After upgrading from a previous npm global install to OpenClaw 2026.4.x (tested on 2026.4.3 and 2026.4.5), every CLI command fails with `Cannot find module` for channel plugin dependencies (`@slack/logger`, `@buape/carbon`, `@larksuiteoapi/node-sdk`, `grammy`), because the `packageManager: pnpm` declaration causes npm to silently drop these packages during global install hoisting.
The newly release minimax-m2.7 is added to both opencode zen and opencode go it is not added in the openclaw models list
CPU Intel(R) Core(TM) i7-4900MQ CPU @ 2.80GHz 2.80 GHz RAM 32.0 GB (31.7 GB 可用) 238 GB SSD SAMSUNG MZ7LN256HCHP-000L7, 1.82 TB HDD ST2000LM015-2E8174 NVIDIA Quadro K3100M (4 GB), Intel(R) HD Graphics 4600 (113 MB) Windows 10 专业版 VERSION:22H2 OS Internal version 19045.6466 I normally use the laptop with the AC power adapter plugged in, and OpenClaw runs completely fine. But as soon as I unplug the adapter and switch to battery power, the OpenClaw Gateway daemon crashes every single time — the failure rate is 100%.
When using moonshotai/kimi-k2.5 (via NVIDIA NIM, api: openai-completions) as the primary agent in OpenClaw 2026.3.11, the agent completely fails to parse and execute any tools (including read, exec, web_search, etc.).
When enabling `VLLM_USE_V2_MODEL_RUNNER=1` with Qwen3.5 models (`Qwen3_5ForConditionalGeneration` / `Qwen3_5ForCausalLM`), the engine crashes during KV cache initialization with an `AssertionError` in `_reshape_kv_cache`.
If a tool with result_as_answer=True is given to agent, Agent ignores the success of tool and make the tool output it's own, which shouldn't happen. result_as_answer=True should work for only successful tool calls ,This essentially removing the capability of agent reflecting on it's output
it is not picking gpt-5.4-mini from openai-codex
The `openclaw-lark` plugin restricts Feishu (Lark) User Access Token (UAT) operations to only the **App Owner**. This means in a multi-user deployment where the bot serves an entire organization, non-owner employees cannot: 1. **Initiate OAuth authorization** — `oauth.js` checks `assertOwnerAccessStrict()` and returns `permission_denied` for non-owners 2. **Call any Feishu user API** — `tool-client.js` calls `assertOwnerAccessStrict()` in `invokeAsUser()`, blocking all UAT-based tools (calendar, tasks, bitable, IM, etc.) for non-owner users