How to Set the Context Length for Ollama with OpenClaude

Set the OLLAMA_MAX_CONTEXT environment variable to 32768 or add max_context: 32768 to ~/.ollama/ollama.yaml, then restart the Ollama service to meet OpenClaude's default token requirement.

OpenClaude assumes a 32,768-token context window when routing chat requests to local Ollama models. If your Ollama instance maintains the default 2,048-token limit, OpenClaude displays a warning because long conversations may be truncated. Configuring Ollama to use the expected context length eliminates this warning and ensures full message history retention.

Why OpenClaude Requires 32,768 Tokens

According to the source code in src/utils/statusNoticeLocalModel.ts, OpenClaude automatically checks the active profile provider and generates a warning when it detects Ollama:

if (activeProfile?.provider?.id === 'ollama') {
  const ollamaContextContributor = await getOllamaContextInfo();
  if (ollamaContextContributor) {
    warnings.push({
      id: 'ollama_context_length',
      title: 'Ollama Context Length',
      description:
        'OpenClaude requests 32768 tokens for Ollama chats. If `ollama ps` keeps showing a smaller CONTEXT after a new request, restart Ollama and verify with `ollama ps`.',
      ...ollamaContextContributor,
    });
  }
}

This warning appears because OpenClaude sends requests assuming a 32,768-token buffer. If Ollama's actual context is smaller, the model discards older messages, leading to incomplete context awareness.

Checking Your Current Ollama Context Size

Before adjusting settings, verify your current context size using the ollama ps command. The getOllamaContextInfo function in src/utils/ollamaContext.ts parses this output to extract the CONTEXT value:

export const getOllamaContextInfo = async (): Promise<LocalModelWarningContributor | null> => {
  try {
    const { stdout } = await execFilePromise('ollama', ['ps']);
    const match = stdout.match(/\bCONTEXT\b\s+(\d+)/i);
    if (match) {
      const context = Number(match[1]);
      return {
        description: `Current Ollama context: ${context} tokens.`,
        details: { context },
      };
    }
  } catch (e) {
    // If ollama is not installed or command fails, silently ignore.
  }
  return null;
};

Run the following command to see your current allocation:

ollama ps

Look for the CONTEXT column. If it shows a value lower than 32768, you need to increase the limit.

Configuring the Ollama Context Length

You can set the maximum context length via environment variables or the Ollama configuration file. Both methods require restarting the Ollama service to take effect.

Method 1: Environment Variable

Set OLLAMA_MAX_CONTEXT before starting the Ollama server:

export OLLAMA_MAX_CONTEXT=32768
ollama serve &

This approach is ideal for temporary testing or containerized deployments where you control the startup environment.

Method 2: Configuration File

For persistent configuration, edit or create ~/.ollama/ollama.yaml:

host: 127.0.0.1
port: 11434
max_context: 32768

After saving the file, restart Ollama:

pkill -f ollama
ollama serve &

Verifying the Configuration

Once restarted, confirm the new context size is active:

ollama ps

The output should display CONTEXT 32768. When you next start a chat session in OpenClaude, the warning in makeLocalModelWarnings (located in src/utils/statusNoticeLocalModel.ts) will no longer trigger because getOllamaContextInfo will report the matching token count.

Summary

• OpenClaude requests 32,768 tokens for all Ollama conversations, as defined in the warning description within src/utils/statusNoticeLocalModel.ts.
• The utility function getOllamaContextInfo in src/utils/ollamaContext.ts executes ollama ps to detect the current context size at runtime.
• Set OLLAMA_MAX_CONTEXT=32768 as an environment variable or add max_context: 32768 to ~/.ollama/ollama.yaml to meet this requirement.
• Always restart the Ollama service after changing configuration values.
• Verify the setting using ollama ps before starting OpenClaude sessions.

Frequently Asked Questions

What context length does OpenClaude expect for Ollama?

OpenClaude expects a 32,768-token context window for Ollama chats. This value is hardcoded in the warning description within src/utils/statusNoticeLocalModel.ts and reflects the default buffer size used when constructing chat requests.

How do I check if my Ollama context is set correctly?

Run ollama ps in your terminal and inspect the CONTEXT column. If it displays 32768, your configuration is correct. Internally, OpenClaude runs the same command via the getOllamaContextInfo function in src/utils/ollamaContext.ts to validate your setup automatically.

Why does OpenClaude show a context length warning even after I changed the setting?

The Ollama service loads configuration only at startup. If you edited ~/.ollama/ollama.yaml or exported the environment variable without restarting the process, the old context limit remains active. Stop all Ollama processes with pkill -f ollama and relaunch with ollama serve to apply the new context length.

Can I use a context length different from 32768 tokens?

While Ollama supports arbitrary context sizes via OLLAMA_MAX_CONTEXT, OpenClaude specifically warns when the value differs from 32,768 tokens. Using a smaller value may cause conversation truncation, while larger values consume more VRAM without benefitting OpenClaude's current implementation.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}