Setting up Ollama with Openclaw on Linux

[swift oracle]

I'm trying to get started with an Openclaw instance on a Pop_OS! Linux PC, that I connected with Ollama. Currently I have Ollama running Openclaw, and I can connect to the local hosted UI using http://localhost:18789. I've set the primary agent as ollama/glm-5.1:cloud. When I try to chat with the model in the local openclaw dashboard, I get an empty loading message and eventually it stops trying to render any message, and just shows: "Assistant 12:45 PM glm-5.1:cloud".

Here's a key-redacted section of my openclaw.json file:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "ollama/glm-5.1:cloud"
      },
      "models": {
        "google/gemini-3.1-pro-preview": {}
      },
      "workspace": "/home/nathansmith/.openclaw/workspace"
    },
    "list": [
      {
        "id": "main",
        "model": "ollama/glm-5.1:cloud"
      }
    ]
  },
  ...
  "models": {
    "providers": {
      "ollama": {
        "api": "ollama",
        "apiKey": "ollama-local",
        "baseUrl": "http://127.0.0.1:11434",
        "models": [
          {
            "contextWindow": 202752,
            "cost": {
              "cacheRead": 0,
              "cacheWrite": 0,
              "input": 0,
              "output": 0
            },
            "id": "glm-5.1:cloud",
            "input": [
              "text"
            ],
            "name": "glm-5.1:cloud",
            "reasoning": true
          }
        ]
      }
    }
  },
  ...

[rapid epoch]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[pearl nexus]

First, can you paste openclaw --version?

I think I see 2 likely issues in that config:

1. agents.defaults.models is an allowlist
   Right now you only allowlisted:

   "google/gemini-3.1-pro-preview": {}
   

   so add your Ollama model there too, or remove the block entirely.

2. "reasoning": true on glm-5.1:cloud is very suspicious
   For reasoning-capable models, OpenClaw defaults thinking to low, and the web chat hides reasoning by default. That can produce exactly what you described: a blank assistant bubble with just the model label.

I’d try this first:

{
  agents: {
    defaults: {
      model: { primary: "ollama/glm-5.1:cloud" },
      thinkingDefault: "off",
      models: {
        "ollama/glm-5.1:cloud": {}
      }
    },
    list: [
      { id: "main", model: "ollama/glm-5.1:cloud" }
    ]
  },
  models: {
    providers: {

      ollama: {
        api: "ollama",
        apiKey: "ollama-local",
        baseUrl: "http://127.0.0.1:11434",
        models: [
          {
            id: "glm-5.1:cloud",
            name: "glm-5.1:cloud",
            input: ["text"],
            contextWindow: 202752,
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 }
          }
        ]
      }
    }

  }
}

So, specifically:

• remove reasoning: true (or set it to false)
• add "ollama/glm-5.1:cloud" to agents.defaults.models, or remove that allowlist
• restart:

  openclaw gateway restart
  

If it still blanks after that, paste:

openclaw status --all

openclaw logs --follow

and send one test message while logs are open.

Refs:

• https://docs.openclaw.ai/help/faq
• https://docs.openclaw.ai/providers/ollama
• https://docs.openclaw.ai/tools/thinking

If this fixes it, please right-click the solution message → Apps → ✅ Mark Solution so Answer Overflow can index it.

[swift oracle]

Openclaw version returns with OpenClaw 2026.4.15 (041266a).

I've made the requested changes to openclaw.json, still see the same chat issue. When I run openclaw status --all, I see the following in the diagnosis section:

Diagnosis (read-only)

Gateway connection details:
  Gateway target: ws://127.0.0.1:18789
  Source: local loopback
  Config: /home/nathansmith/.openclaw/openclaw.json
  Bind: loopback

✓ Config: /home/nathansmith/.openclaw/openclaw.json
✓ Secret diagnostics (0)
✓ Restart sentinel: none
! Port 18789
  Port 18789 is already in use.
  - pid 20436 nathansmith: openclaw-gateway (127.0.0.1:18789)
  - Gateway already running locally. Stop it (openclaw gateway stop) or use a different port.
! Tailscale: off · unknown
✓ Skills: 8 eligible · 0 missing · /home/nathansmith/.openclaw/workspace
✓ Plugin compatibility (none)
✓ Channel issues (none)

Ollama is running openclaw's server in that port, hence the "already in use" alert on port 18789.

[pearl nexus]

Yep, that Port 18789 is already in use line is not the problem here, it just means the OpenClaw gateway is already running on the expected port.

Since the config change didn’t fix it, the next step is to split UI problem vs model/backend problem.

Please try these:

1. In the dashboard session, force:

   • Thinking = off
   • Reasoning visibility = on

   Or send:

   /think off
   /reasoning on
   

   Then ask:

   Reply with exactly GLM-OK.

2. Run the same model locally from CLI, bypassing the web UI:

   openclaw agent --local --agent main --thinking off -m "Reply with exactly GLM-OK." --json
   

3. If that is still blank, run:

   openclaw logs --follow
   

   then send one test message in the dashboard and paste the relevant log lines.

Why I’m asking for this:

• if CLI works, the model/backend is fine and the issue is likely the web chat session/UI state

• if CLI is also blank, then it’s a provider/model runtime issue

One more thing: you’re on 2026.4.15, and there have been a bunch of Ollama/thinking fixes after that, so updating is worth trying if the CLI test still looks weird:
https://docs.openclaw.ai/install/updating

Also useful refs:

• https://docs.openclaw.ai/providers/ollama
• https://docs.openclaw.ai/tools/thinking
• https://docs.openclaw.ai/help/troubleshooting

Paste the openclaw agent ... --json result first, that should tell us a lot fast.

[swift oracle]

I've made the session updates in the dashboard on Control > Sessions, for the only entry there, agent:main:main, reasoning -> on, thinking -> off.

Asking in the chat again, I see the message is Queued (1)

                    Reply with exactly GLM-OK.

Running locally by CLI (step 2), I get this in payloads:

"payloads": [
    {
      "text": "403 {\"error\":\"model is experiencing high volume. while capacity is being added, a subscription is required for access: https://ollama.com/upgrade (ref: 9a130925-8a11-4e2e-9d27-d706a2b6695d)\"}",
      "mediaUrl": null
    }
  ]

I probably just need to start a subscription and try again

[pearl nexus]

Yep, that’s the real issue. Nice, we found it.

This means OpenClaw itself is basically fine, and the failure is upstream from Ollama cloud:

403 ... model is experiencing high volume ... a subscription is required for access

So for ollama/glm-5.1:cloud, your next options are:

1. Use Ollama cloud with a signed-in/subscribed account

   • make sure ollama signin is done on that host
   • if Ollama is rate-limiting/free-tier gating that model, a subscription may indeed be required

2. Switch to a local Ollama model instead

   • check what you already have:

ollama list
openclaw models list --provider ollama
```

• then set one of those local models as primary

The Queued (1) in the dashboard was probably just the request sitting behind a blocked upstream run, not a separate root cause.

Docs:

• https://docs.openclaw.ai/providers/ollama
• https://docs.openclaw.ai/help/faq

If you want, paste your ollama list output and I can help you turn that into the exact openclaw.json model entry.
If this solved the mystery, please right-click the solution message → Apps → ✅ Mark Solution so Answer Overflow can index it.

[swift oracle]

I paid for a subscription. Istill see messages in the dashboard showing up in the queue, but in the CLI I now get the payload:

  "payloads": [
    {
      "text": "GLM-OK",
      "mediaUrl": null
    }
  ],