Bug type
Regression (worked before, now fails)
Beta release blocker
No
Summary
Upgraded from version 2026.3.31 (maybe) to 2026.5.19, OpenClaw cannot connect to local Ollama and reports: No API key found for provider "ollama"
Steps to reproduce
- Run Ollama on the host/root system.
- Run OpenClaw inside the Orb virtual machine. Select a local Ollama model without environment values.
- Upgrade OpenClaw from a version before 2026.3.31 to 2026.5.19.
- Select a local Ollama model and send a message.
- Try running auth login again.
Expected behavior
- OpenClaw inside the Orb VM should connect to the host Ollama instance normally.
- Local Ollama models should work without requiring a real API key or auth profile.
- Behavior should remain the same as before the upgrade.
Actual behavior
- OpenClaw reports: No API key found for provider "ollama"
- No Ollama auth-profiles are available.
- Running auth login again still cannot connect to Ollama.
- Local oMLX models work normally from the same OpenClaw instance.
- Everything worked correctly before upgrading.
OpenClaw version
2026.5.19
Operating system
macOS 15.7 host, Ubuntu 24.04 orb virtual
Install method
pnpm global from source
Model
qwen3.5:27b
Provider / routing chain
openclaw -> ollama -> qwen3.5:27b
Additional provider/model setup details
No response
Logs, screenshots, and evidence
model config
"ollama": {
"apiKey": "ollama-local",
"baseUrl": "http://host.orb.internal:11434",
"api": "ollama",
"models": [
{
"id": "kimi-k2.5:cloud",
"name": "kimi-k2.5:cloud",
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"maxTokens": 8192,
"params": {
"num_ctx": 128000
},
"api": "ollama",
"reasoning": false,
"input": [
"text"
]
},
{
"id": "minimax-m2.5:cloud",
"name": "minimax-m2.5:cloud",
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 128000,
"api": "ollama",
"input": [
"text"
],
"reasoning": false,
"maxTokens": 8192,
"params": {
"num_ctx": 128000
}
},
{
"id": "glm-5:cloud",
"name": "glm-5:cloud",
"cost": {
"input": 0,
"output": 0,
"cacheWrite": 0,
"cacheRead": 0
},
"contextWindow": 128000,
"reasoning": false,
"api": "ollama",
"params": {
"num_ctx": 128000
},
"input": [
"text"
],
"maxTokens": 8192
},
{
"id": "qwen3.5:27b",
"name": "qwen3.5:27b",
"cost": {
"input": 0,
"output": 0,
"cacheWrite": 0,
"cacheRead": 0
},
"contextWindow": 262144,
"params": {
"num_ctx": 262144
},
"maxTokens": 8192,
"reasoning": false,
"api": "ollama",
"input": [
"text"
]
},
{
"id": "glm-4.7-flash:latest",
"name": "glm-4.7-flash:latest",
"cost": {
"input": 0,
"output": 0,
"cacheWrite": 0,
"cacheRead": 0
},
"contextWindow": 202752,
"params": {
"num_ctx": 202752
},
"maxTokens": 8192,
"input": [
"text"
],
"reasoning": false
}
]
},
"omlx-local-mac": {
"baseUrl": "http://host.orb.internal:8000/v1",
"apiKey": "1234",
"api": "openai-completions",
"models": [
{
"id": "Qwen3.5-27B-4bit",
"name": "qwen3.5:27B",
"input": [
"text"
]
}
]
}
openclaw logs
06:08:29+00:00 error Embedded agent failed before reply: No API key found for provider "ollama". Auth store: /home/nirvanach7/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/nirvanach7/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir. | No API key found for provider "ollama". Auth store: /home/nirvanach7/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/nirvanach7/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir.
auth login
nirvanach7@openclaw:~/.openclaw$ openclaw models auth login --provider ollama
🦞 OpenClaw 2026.5.19 (a185ca2) — Your second brain, except this one actually remembers where you left things.
│
◇ Ollama mode
│ Cloud + Local
│
◇ Ollama base URL
│ http://host.orb.internal:11434
│
◇ Ollama ─────────────────────────────────────────────────────────╮
│ │
│ Ollama could not be reached at http://host.orb.internal:11434. │
│ Download it at https://ollama.com/download │
│ │
│ Start Ollama and re-run setup. │
│ │
├──────────────────────────────────────────────────────────────────╯
WizardCancelledError: Ollama not reachable
nirvanach7@openclaw:~/.openclaw$ curl http://host.orb.internal:11434
Ollama is running
nirvanach7@openclaw:~/.openclaw$ curl http://host.orb.internal:11434/api/generate -d '{
"model": "qwen3.5:27b",
"prompt": "test"
}'
{"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.276034Z","response":"","thinking":"Okay","done":false}
{"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.423831Z","response":"","thinking":",","done":false}
{"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.563806Z","response":"","thinking":" the","done":false}
{"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.70267Z","response":"","thinking":" user","done":false}
{"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.842449Z","response":"","thinking":" just","done":false}
{"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.983375Z","response":"","thinking":" sent","done":false}
^CImpact and severity
Affected:Users running OpenClaw in an Orb VM connecting to Ollama running on the host/root machine.
Severity:Blocks use of local Ollama models (cannot send messages).
Frequency:Always occurs after upgrading.
Consequence:Ollama models become unusable in OpenClaw (shows “No API key found for provider "ollama"”), while other providers like oMLX still work.
Additional information
Regression observed: Ollama worked correctly before the upgrade.
Last known good version: before 2026.3.31 (exact version uncertain)
First known bad version: 2026.3.31–2026.5.19 range after upgrade
Ollama runs on host/root system, while OpenClaw runs inside an Orb virtual machine.
oMLX provider continues to work normally in the same environment.
Bug type
Regression (worked before, now fails)
Beta release blocker
No
Summary
Upgraded from version 2026.3.31 (maybe) to 2026.5.19, OpenClaw cannot connect to local Ollama and reports: No API key found for provider "ollama"
Steps to reproduce
Expected behavior
Actual behavior
OpenClaw version
2026.5.19
Operating system
macOS 15.7 host, Ubuntu 24.04 orb virtual
Install method
pnpm global from source
Model
qwen3.5:27b
Provider / routing chain
openclaw -> ollama -> qwen3.5:27b
Additional provider/model setup details
No response
Logs, screenshots, and evidence
model config "ollama": { "apiKey": "ollama-local", "baseUrl": "http://host.orb.internal:11434", "api": "ollama", "models": [ { "id": "kimi-k2.5:cloud", "name": "kimi-k2.5:cloud", "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 128000, "maxTokens": 8192, "params": { "num_ctx": 128000 }, "api": "ollama", "reasoning": false, "input": [ "text" ] }, { "id": "minimax-m2.5:cloud", "name": "minimax-m2.5:cloud", "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 }, "contextWindow": 128000, "api": "ollama", "input": [ "text" ], "reasoning": false, "maxTokens": 8192, "params": { "num_ctx": 128000 } }, { "id": "glm-5:cloud", "name": "glm-5:cloud", "cost": { "input": 0, "output": 0, "cacheWrite": 0, "cacheRead": 0 }, "contextWindow": 128000, "reasoning": false, "api": "ollama", "params": { "num_ctx": 128000 }, "input": [ "text" ], "maxTokens": 8192 }, { "id": "qwen3.5:27b", "name": "qwen3.5:27b", "cost": { "input": 0, "output": 0, "cacheWrite": 0, "cacheRead": 0 }, "contextWindow": 262144, "params": { "num_ctx": 262144 }, "maxTokens": 8192, "reasoning": false, "api": "ollama", "input": [ "text" ] }, { "id": "glm-4.7-flash:latest", "name": "glm-4.7-flash:latest", "cost": { "input": 0, "output": 0, "cacheWrite": 0, "cacheRead": 0 }, "contextWindow": 202752, "params": { "num_ctx": 202752 }, "maxTokens": 8192, "input": [ "text" ], "reasoning": false } ] }, "omlx-local-mac": { "baseUrl": "http://host.orb.internal:8000/v1", "apiKey": "1234", "api": "openai-completions", "models": [ { "id": "Qwen3.5-27B-4bit", "name": "qwen3.5:27B", "input": [ "text" ] } ] } openclaw logs 06:08:29+00:00 error Embedded agent failed before reply: No API key found for provider "ollama". Auth store: /home/nirvanach7/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/nirvanach7/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir. | No API key found for provider "ollama". Auth store: /home/nirvanach7/.openclaw/agents/main/agent/auth-profiles.json (agentDir: /home/nirvanach7/.openclaw/agents/main/agent). Configure auth for this agent (openclaw agents add <id>) or copy only portable static auth profiles from the main agentDir. auth login nirvanach7@openclaw:~/.openclaw$ openclaw models auth login --provider ollama 🦞 OpenClaw 2026.5.19 (a185ca2) — Your second brain, except this one actually remembers where you left things. │ ◇ Ollama mode │ Cloud + Local │ ◇ Ollama base URL │ http://host.orb.internal:11434 │ ◇ Ollama ─────────────────────────────────────────────────────────╮ │ │ │ Ollama could not be reached at http://host.orb.internal:11434. │ │ Download it at https://ollama.com/download │ │ │ │ Start Ollama and re-run setup. │ │ │ ├──────────────────────────────────────────────────────────────────╯ WizardCancelledError: Ollama not reachable nirvanach7@openclaw:~/.openclaw$ curl http://host.orb.internal:11434 Ollama is running nirvanach7@openclaw:~/.openclaw$ curl http://host.orb.internal:11434/api/generate -d '{ "model": "qwen3.5:27b", "prompt": "test" }' {"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.276034Z","response":"","thinking":"Okay","done":false} {"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.423831Z","response":"","thinking":",","done":false} {"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.563806Z","response":"","thinking":" the","done":false} {"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.70267Z","response":"","thinking":" user","done":false} {"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.842449Z","response":"","thinking":" just","done":false} {"model":"qwen3.5:27b","created_at":"2026-05-21T07:46:41.983375Z","response":"","thinking":" sent","done":false} ^CImpact and severity
Affected:Users running OpenClaw in an Orb VM connecting to Ollama running on the host/root machine.
Severity:Blocks use of local Ollama models (cannot send messages).
Frequency:Always occurs after upgrading.
Consequence:Ollama models become unusable in OpenClaw (shows “No API key found for provider "ollama"”), while other providers like oMLX still work.
Additional information
Regression observed: Ollama worked correctly before the upgrade.
Last known good version: before 2026.3.31 (exact version uncertain)
First known bad version: 2026.3.31–2026.5.19 range after upgrade
Ollama runs on host/root system, while OpenClaw runs inside an Orb virtual machine.
oMLX provider continues to work normally in the same environment.